Loading...
1/*
2 * net/tipc/socket.c: TIPC socket API
3 *
4 * Copyright (c) 2001-2007, 2012-2019, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6 * Copyright (c) 2020-2021, Red Hat Inc
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions are met:
11 *
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the names of the copyright holders nor the names of its
18 * contributors may be used to endorse or promote products derived from
19 * this software without specific prior written permission.
20 *
21 * Alternatively, this software may be distributed under the terms of the
22 * GNU General Public License ("GPL") version 2 as published by the Free
23 * Software Foundation.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
36 */
37
38#include <linux/rhashtable.h>
39#include <linux/sched/signal.h>
40#include <trace/events/sock.h>
41
42#include "core.h"
43#include "name_table.h"
44#include "node.h"
45#include "link.h"
46#include "name_distr.h"
47#include "socket.h"
48#include "bcast.h"
49#include "netlink.h"
50#include "group.h"
51#include "trace.h"
52
53#define NAGLE_START_INIT 4
54#define NAGLE_START_MAX 1024
55#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
56#define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
57#define TIPC_MAX_PORT 0xffffffff
58#define TIPC_MIN_PORT 1
59#define TIPC_ACK_RATE 4 /* ACK at 1/4 of rcv window size */
60
61enum {
62 TIPC_LISTEN = TCP_LISTEN,
63 TIPC_ESTABLISHED = TCP_ESTABLISHED,
64 TIPC_OPEN = TCP_CLOSE,
65 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
66 TIPC_CONNECTING = TCP_SYN_SENT,
67};
68
69struct sockaddr_pair {
70 struct sockaddr_tipc sock;
71 struct sockaddr_tipc member;
72};
73
74/**
75 * struct tipc_sock - TIPC socket structure
76 * @sk: socket - interacts with 'port' and with user via the socket API
77 * @max_pkt: maximum packet size "hint" used when building messages sent by port
78 * @maxnagle: maximum size of msg which can be subject to nagle
79 * @portid: unique port identity in TIPC socket hash table
80 * @phdr: preformatted message header used when sending messages
81 * @cong_links: list of congested links
82 * @publications: list of publications for port
83 * @pub_count: total # of publications port has made during its lifetime
84 * @conn_timeout: the time we can wait for an unresponded setup request
85 * @probe_unacked: probe has not received ack yet
86 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
87 * @cong_link_cnt: number of congested links
88 * @snt_unacked: # messages sent by socket, and not yet acked by peer
89 * @snd_win: send window size
90 * @peer_caps: peer capabilities mask
91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
92 * @rcv_win: receive window size
93 * @peer: 'connected' peer for dgram/rdm
94 * @node: hash table node
95 * @mc_method: cookie for use between socket and broadcast layer
96 * @rcu: rcu struct for tipc_sock
97 * @group: TIPC communications group
98 * @oneway: message count in one direction (FIXME)
99 * @nagle_start: current nagle value
100 * @snd_backlog: send backlog count
101 * @msg_acc: messages accepted; used in managing backlog and nagle
102 * @pkt_cnt: TIPC socket packet count
103 * @expect_ack: whether this TIPC socket is expecting an ack
104 * @nodelay: setsockopt() TIPC_NODELAY setting
105 * @group_is_open: TIPC socket group is fully open (FIXME)
106 * @published: true if port has one or more associated names
107 * @conn_addrtype: address type used when establishing connection
108 */
109struct tipc_sock {
110 struct sock sk;
111 u32 max_pkt;
112 u32 maxnagle;
113 u32 portid;
114 struct tipc_msg phdr;
115 struct list_head cong_links;
116 struct list_head publications;
117 u32 pub_count;
118 atomic_t dupl_rcvcnt;
119 u16 conn_timeout;
120 bool probe_unacked;
121 u16 cong_link_cnt;
122 u16 snt_unacked;
123 u16 snd_win;
124 u16 peer_caps;
125 u16 rcv_unacked;
126 u16 rcv_win;
127 struct sockaddr_tipc peer;
128 struct rhash_head node;
129 struct tipc_mc_method mc_method;
130 struct rcu_head rcu;
131 struct tipc_group *group;
132 u32 oneway;
133 u32 nagle_start;
134 u16 snd_backlog;
135 u16 msg_acc;
136 u16 pkt_cnt;
137 bool expect_ack;
138 bool nodelay;
139 bool group_is_open;
140 bool published;
141 u8 conn_addrtype;
142};
143
144static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
145static void tipc_data_ready(struct sock *sk);
146static void tipc_write_space(struct sock *sk);
147static void tipc_sock_destruct(struct sock *sk);
148static int tipc_release(struct socket *sock);
149static void tipc_sk_timeout(struct timer_list *t);
150static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
151static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
152static int tipc_sk_leave(struct tipc_sock *tsk);
153static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
154static int tipc_sk_insert(struct tipc_sock *tsk);
155static void tipc_sk_remove(struct tipc_sock *tsk);
156static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
157static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
158static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
159static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
160
161static const struct proto_ops packet_ops;
162static const struct proto_ops stream_ops;
163static const struct proto_ops msg_ops;
164static struct proto tipc_proto;
165static const struct rhashtable_params tsk_rht_params;
166
167static u32 tsk_own_node(struct tipc_sock *tsk)
168{
169 return msg_prevnode(&tsk->phdr);
170}
171
172static u32 tsk_peer_node(struct tipc_sock *tsk)
173{
174 return msg_destnode(&tsk->phdr);
175}
176
177static u32 tsk_peer_port(struct tipc_sock *tsk)
178{
179 return msg_destport(&tsk->phdr);
180}
181
182static bool tsk_unreliable(struct tipc_sock *tsk)
183{
184 return msg_src_droppable(&tsk->phdr) != 0;
185}
186
187static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
188{
189 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
190}
191
192static bool tsk_unreturnable(struct tipc_sock *tsk)
193{
194 return msg_dest_droppable(&tsk->phdr) != 0;
195}
196
197static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
198{
199 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
200}
201
202static int tsk_importance(struct tipc_sock *tsk)
203{
204 return msg_importance(&tsk->phdr);
205}
206
207static struct tipc_sock *tipc_sk(const struct sock *sk)
208{
209 return container_of(sk, struct tipc_sock, sk);
210}
211
212int tsk_set_importance(struct sock *sk, int imp)
213{
214 if (imp > TIPC_CRITICAL_IMPORTANCE)
215 return -EINVAL;
216 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
217 return 0;
218}
219
220static bool tsk_conn_cong(struct tipc_sock *tsk)
221{
222 return tsk->snt_unacked > tsk->snd_win;
223}
224
225static u16 tsk_blocks(int len)
226{
227 return ((len / FLOWCTL_BLK_SZ) + 1);
228}
229
230/* tsk_blocks(): translate a buffer size in bytes to number of
231 * advertisable blocks, taking into account the ratio truesize(len)/len
232 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
233 */
234static u16 tsk_adv_blocks(int len)
235{
236 return len / FLOWCTL_BLK_SZ / 4;
237}
238
239/* tsk_inc(): increment counter for sent or received data
240 * - If block based flow control is not supported by peer we
241 * fall back to message based ditto, incrementing the counter
242 */
243static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
244{
245 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
246 return ((msglen / FLOWCTL_BLK_SZ) + 1);
247 return 1;
248}
249
250/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
251 */
252static void tsk_set_nagle(struct tipc_sock *tsk)
253{
254 struct sock *sk = &tsk->sk;
255
256 tsk->maxnagle = 0;
257 if (sk->sk_type != SOCK_STREAM)
258 return;
259 if (tsk->nodelay)
260 return;
261 if (!(tsk->peer_caps & TIPC_NAGLE))
262 return;
263 /* Limit node local buffer size to avoid receive queue overflow */
264 if (tsk->max_pkt == MAX_MSG_SIZE)
265 tsk->maxnagle = 1500;
266 else
267 tsk->maxnagle = tsk->max_pkt;
268}
269
270/**
271 * tsk_advance_rx_queue - discard first buffer in socket receive queue
272 * @sk: network socket
273 *
274 * Caller must hold socket lock
275 */
276static void tsk_advance_rx_queue(struct sock *sk)
277{
278 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
279 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
280}
281
282/* tipc_sk_respond() : send response message back to sender
283 */
284static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
285{
286 u32 selector;
287 u32 dnode;
288 u32 onode = tipc_own_addr(sock_net(sk));
289
290 if (!tipc_msg_reverse(onode, &skb, err))
291 return;
292
293 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
294 dnode = msg_destnode(buf_msg(skb));
295 selector = msg_origport(buf_msg(skb));
296 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
297}
298
299/**
300 * tsk_rej_rx_queue - reject all buffers in socket receive queue
301 * @sk: network socket
302 * @error: response error code
303 *
304 * Caller must hold socket lock
305 */
306static void tsk_rej_rx_queue(struct sock *sk, int error)
307{
308 struct sk_buff *skb;
309
310 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
311 tipc_sk_respond(sk, skb, error);
312}
313
314static bool tipc_sk_connected(const struct sock *sk)
315{
316 return READ_ONCE(sk->sk_state) == TIPC_ESTABLISHED;
317}
318
319/* tipc_sk_type_connectionless - check if the socket is datagram socket
320 * @sk: socket
321 *
322 * Returns true if connection less, false otherwise
323 */
324static bool tipc_sk_type_connectionless(struct sock *sk)
325{
326 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
327}
328
329/* tsk_peer_msg - verify if message was sent by connected port's peer
330 *
331 * Handles cases where the node's network address has changed from
332 * the default of <0.0.0> to its configured setting.
333 */
334static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
335{
336 struct sock *sk = &tsk->sk;
337 u32 self = tipc_own_addr(sock_net(sk));
338 u32 peer_port = tsk_peer_port(tsk);
339 u32 orig_node, peer_node;
340
341 if (unlikely(!tipc_sk_connected(sk)))
342 return false;
343
344 if (unlikely(msg_origport(msg) != peer_port))
345 return false;
346
347 orig_node = msg_orignode(msg);
348 peer_node = tsk_peer_node(tsk);
349
350 if (likely(orig_node == peer_node))
351 return true;
352
353 if (!orig_node && peer_node == self)
354 return true;
355
356 if (!peer_node && orig_node == self)
357 return true;
358
359 return false;
360}
361
362/* tipc_set_sk_state - set the sk_state of the socket
363 * @sk: socket
364 *
365 * Caller must hold socket lock
366 *
367 * Returns 0 on success, errno otherwise
368 */
369static int tipc_set_sk_state(struct sock *sk, int state)
370{
371 int oldsk_state = sk->sk_state;
372 int res = -EINVAL;
373
374 switch (state) {
375 case TIPC_OPEN:
376 res = 0;
377 break;
378 case TIPC_LISTEN:
379 case TIPC_CONNECTING:
380 if (oldsk_state == TIPC_OPEN)
381 res = 0;
382 break;
383 case TIPC_ESTABLISHED:
384 if (oldsk_state == TIPC_CONNECTING ||
385 oldsk_state == TIPC_OPEN)
386 res = 0;
387 break;
388 case TIPC_DISCONNECTING:
389 if (oldsk_state == TIPC_CONNECTING ||
390 oldsk_state == TIPC_ESTABLISHED)
391 res = 0;
392 break;
393 }
394
395 if (!res)
396 sk->sk_state = state;
397
398 return res;
399}
400
401static int tipc_sk_sock_err(struct socket *sock, long *timeout)
402{
403 struct sock *sk = sock->sk;
404 int err = sock_error(sk);
405 int typ = sock->type;
406
407 if (err)
408 return err;
409 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
410 if (sk->sk_state == TIPC_DISCONNECTING)
411 return -EPIPE;
412 else if (!tipc_sk_connected(sk))
413 return -ENOTCONN;
414 }
415 if (!*timeout)
416 return -EAGAIN;
417 if (signal_pending(current))
418 return sock_intr_errno(*timeout);
419
420 return 0;
421}
422
423#define tipc_wait_for_cond(sock_, timeo_, condition_) \
424({ \
425 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
426 struct sock *sk_; \
427 int rc_; \
428 \
429 while ((rc_ = !(condition_))) { \
430 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
431 smp_rmb(); \
432 sk_ = (sock_)->sk; \
433 rc_ = tipc_sk_sock_err((sock_), timeo_); \
434 if (rc_) \
435 break; \
436 add_wait_queue(sk_sleep(sk_), &wait_); \
437 release_sock(sk_); \
438 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
439 sched_annotate_sleep(); \
440 lock_sock(sk_); \
441 remove_wait_queue(sk_sleep(sk_), &wait_); \
442 } \
443 rc_; \
444})
445
446/**
447 * tipc_sk_create - create a TIPC socket
448 * @net: network namespace (must be default network)
449 * @sock: pre-allocated socket structure
450 * @protocol: protocol indicator (must be 0)
451 * @kern: caused by kernel or by userspace?
452 *
453 * This routine creates additional data structures used by the TIPC socket,
454 * initializes them, and links them together.
455 *
456 * Return: 0 on success, errno otherwise
457 */
458static int tipc_sk_create(struct net *net, struct socket *sock,
459 int protocol, int kern)
460{
461 const struct proto_ops *ops;
462 struct sock *sk;
463 struct tipc_sock *tsk;
464 struct tipc_msg *msg;
465
466 /* Validate arguments */
467 if (unlikely(protocol != 0))
468 return -EPROTONOSUPPORT;
469
470 switch (sock->type) {
471 case SOCK_STREAM:
472 ops = &stream_ops;
473 break;
474 case SOCK_SEQPACKET:
475 ops = &packet_ops;
476 break;
477 case SOCK_DGRAM:
478 case SOCK_RDM:
479 ops = &msg_ops;
480 break;
481 default:
482 return -EPROTOTYPE;
483 }
484
485 /* Allocate socket's protocol area */
486 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
487 if (sk == NULL)
488 return -ENOMEM;
489
490 tsk = tipc_sk(sk);
491 tsk->max_pkt = MAX_PKT_DEFAULT;
492 tsk->maxnagle = 0;
493 tsk->nagle_start = NAGLE_START_INIT;
494 INIT_LIST_HEAD(&tsk->publications);
495 INIT_LIST_HEAD(&tsk->cong_links);
496 msg = &tsk->phdr;
497
498 /* Finish initializing socket data structures */
499 sock->ops = ops;
500 sock_init_data(sock, sk);
501 tipc_set_sk_state(sk, TIPC_OPEN);
502 if (tipc_sk_insert(tsk)) {
503 sk_free(sk);
504 pr_warn("Socket create failed; port number exhausted\n");
505 return -EINVAL;
506 }
507
508 /* Ensure tsk is visible before we read own_addr. */
509 smp_mb();
510
511 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
512 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
513
514 msg_set_origport(msg, tsk->portid);
515 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
516 sk->sk_shutdown = 0;
517 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
518 sk->sk_rcvbuf = READ_ONCE(sysctl_tipc_rmem[1]);
519 sk->sk_data_ready = tipc_data_ready;
520 sk->sk_write_space = tipc_write_space;
521 sk->sk_destruct = tipc_sock_destruct;
522 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
523 tsk->group_is_open = true;
524 atomic_set(&tsk->dupl_rcvcnt, 0);
525
526 /* Start out with safe limits until we receive an advertised window */
527 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
528 tsk->rcv_win = tsk->snd_win;
529
530 if (tipc_sk_type_connectionless(sk)) {
531 tsk_set_unreturnable(tsk, true);
532 if (sock->type == SOCK_DGRAM)
533 tsk_set_unreliable(tsk, true);
534 }
535 __skb_queue_head_init(&tsk->mc_method.deferredq);
536 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
537 return 0;
538}
539
540static void tipc_sk_callback(struct rcu_head *head)
541{
542 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
543
544 sock_put(&tsk->sk);
545}
546
547/* Caller should hold socket lock for the socket. */
548static void __tipc_shutdown(struct socket *sock, int error)
549{
550 struct sock *sk = sock->sk;
551 struct tipc_sock *tsk = tipc_sk(sk);
552 struct net *net = sock_net(sk);
553 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
554 u32 dnode = tsk_peer_node(tsk);
555 struct sk_buff *skb;
556
557 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
558 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
559 !tsk_conn_cong(tsk)));
560
561 /* Push out delayed messages if in Nagle mode */
562 tipc_sk_push_backlog(tsk, false);
563 /* Remove pending SYN */
564 __skb_queue_purge(&sk->sk_write_queue);
565
566 /* Remove partially received buffer if any */
567 skb = skb_peek(&sk->sk_receive_queue);
568 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
569 __skb_unlink(skb, &sk->sk_receive_queue);
570 kfree_skb(skb);
571 }
572
573 /* Reject all unreceived messages if connectionless */
574 if (tipc_sk_type_connectionless(sk)) {
575 tsk_rej_rx_queue(sk, error);
576 return;
577 }
578
579 switch (sk->sk_state) {
580 case TIPC_CONNECTING:
581 case TIPC_ESTABLISHED:
582 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
583 tipc_node_remove_conn(net, dnode, tsk->portid);
584 /* Send a FIN+/- to its peer */
585 skb = __skb_dequeue(&sk->sk_receive_queue);
586 if (skb) {
587 __skb_queue_purge(&sk->sk_receive_queue);
588 tipc_sk_respond(sk, skb, error);
589 break;
590 }
591 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
592 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
593 tsk_own_node(tsk), tsk_peer_port(tsk),
594 tsk->portid, error);
595 if (skb)
596 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
597 break;
598 case TIPC_LISTEN:
599 /* Reject all SYN messages */
600 tsk_rej_rx_queue(sk, error);
601 break;
602 default:
603 __skb_queue_purge(&sk->sk_receive_queue);
604 break;
605 }
606}
607
608/**
609 * tipc_release - destroy a TIPC socket
610 * @sock: socket to destroy
611 *
612 * This routine cleans up any messages that are still queued on the socket.
613 * For DGRAM and RDM socket types, all queued messages are rejected.
614 * For SEQPACKET and STREAM socket types, the first message is rejected
615 * and any others are discarded. (If the first message on a STREAM socket
616 * is partially-read, it is discarded and the next one is rejected instead.)
617 *
618 * NOTE: Rejected messages are not necessarily returned to the sender! They
619 * are returned or discarded according to the "destination droppable" setting
620 * specified for the message by the sender.
621 *
622 * Return: 0 on success, errno otherwise
623 */
624static int tipc_release(struct socket *sock)
625{
626 struct sock *sk = sock->sk;
627 struct tipc_sock *tsk;
628
629 /*
630 * Exit if socket isn't fully initialized (occurs when a failed accept()
631 * releases a pre-allocated child socket that was never used)
632 */
633 if (sk == NULL)
634 return 0;
635
636 tsk = tipc_sk(sk);
637 lock_sock(sk);
638
639 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
640 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
641 sk->sk_shutdown = SHUTDOWN_MASK;
642 tipc_sk_leave(tsk);
643 tipc_sk_withdraw(tsk, NULL);
644 __skb_queue_purge(&tsk->mc_method.deferredq);
645 sk_stop_timer(sk, &sk->sk_timer);
646 tipc_sk_remove(tsk);
647
648 sock_orphan(sk);
649 /* Reject any messages that accumulated in backlog queue */
650 release_sock(sk);
651 tipc_dest_list_purge(&tsk->cong_links);
652 tsk->cong_link_cnt = 0;
653 call_rcu(&tsk->rcu, tipc_sk_callback);
654 sock->sk = NULL;
655
656 return 0;
657}
658
659/**
660 * __tipc_bind - associate or disassociate TIPC name(s) with a socket
661 * @sock: socket structure
662 * @skaddr: socket address describing name(s) and desired operation
663 * @alen: size of socket address data structure
664 *
665 * Name and name sequence binding are indicated using a positive scope value;
666 * a negative scope value unbinds the specified name. Specifying no name
667 * (i.e. a socket address length of 0) unbinds all names from the socket.
668 *
669 * Return: 0 on success, errno otherwise
670 *
671 * NOTE: This routine doesn't need to take the socket lock since it doesn't
672 * access any non-constant socket information.
673 */
674static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
675{
676 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
677 struct tipc_sock *tsk = tipc_sk(sock->sk);
678 bool unbind = false;
679
680 if (unlikely(!alen))
681 return tipc_sk_withdraw(tsk, NULL);
682
683 if (ua->addrtype == TIPC_SERVICE_ADDR) {
684 ua->addrtype = TIPC_SERVICE_RANGE;
685 ua->sr.upper = ua->sr.lower;
686 }
687 if (ua->scope < 0) {
688 unbind = true;
689 ua->scope = -ua->scope;
690 }
691 /* Users may still use deprecated TIPC_ZONE_SCOPE */
692 if (ua->scope != TIPC_NODE_SCOPE)
693 ua->scope = TIPC_CLUSTER_SCOPE;
694
695 if (tsk->group)
696 return -EACCES;
697
698 if (unbind)
699 return tipc_sk_withdraw(tsk, ua);
700 return tipc_sk_publish(tsk, ua);
701}
702
703int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
704{
705 int res;
706
707 lock_sock(sock->sk);
708 res = __tipc_bind(sock, skaddr, alen);
709 release_sock(sock->sk);
710 return res;
711}
712
713static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
714{
715 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
716 u32 atype = ua->addrtype;
717
718 if (alen) {
719 if (!tipc_uaddr_valid(ua, alen))
720 return -EINVAL;
721 if (atype == TIPC_SOCKET_ADDR)
722 return -EAFNOSUPPORT;
723 if (ua->sr.type < TIPC_RESERVED_TYPES) {
724 pr_warn_once("Can't bind to reserved service type %u\n",
725 ua->sr.type);
726 return -EACCES;
727 }
728 }
729 return tipc_sk_bind(sock, skaddr, alen);
730}
731
732/**
733 * tipc_getname - get port ID of socket or peer socket
734 * @sock: socket structure
735 * @uaddr: area for returned socket address
736 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
737 *
738 * Return: 0 on success, errno otherwise
739 *
740 * NOTE: This routine doesn't need to take the socket lock since it only
741 * accesses socket information that is unchanging (or which changes in
742 * a completely predictable manner).
743 */
744static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
745 int peer)
746{
747 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
748 struct sock *sk = sock->sk;
749 struct tipc_sock *tsk = tipc_sk(sk);
750
751 memset(addr, 0, sizeof(*addr));
752 if (peer) {
753 if ((!tipc_sk_connected(sk)) &&
754 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
755 return -ENOTCONN;
756 addr->addr.id.ref = tsk_peer_port(tsk);
757 addr->addr.id.node = tsk_peer_node(tsk);
758 } else {
759 addr->addr.id.ref = tsk->portid;
760 addr->addr.id.node = tipc_own_addr(sock_net(sk));
761 }
762
763 addr->addrtype = TIPC_SOCKET_ADDR;
764 addr->family = AF_TIPC;
765 addr->scope = 0;
766 addr->addr.name.domain = 0;
767
768 return sizeof(*addr);
769}
770
771/**
772 * tipc_poll - read and possibly block on pollmask
773 * @file: file structure associated with the socket
774 * @sock: socket for which to calculate the poll bits
775 * @wait: ???
776 *
777 * Return: pollmask value
778 *
779 * COMMENTARY:
780 * It appears that the usual socket locking mechanisms are not useful here
781 * since the pollmask info is potentially out-of-date the moment this routine
782 * exits. TCP and other protocols seem to rely on higher level poll routines
783 * to handle any preventable race conditions, so TIPC will do the same ...
784 *
785 * IMPORTANT: The fact that a read or write operation is indicated does NOT
786 * imply that the operation will succeed, merely that it should be performed
787 * and will not block.
788 */
789static __poll_t tipc_poll(struct file *file, struct socket *sock,
790 poll_table *wait)
791{
792 struct sock *sk = sock->sk;
793 struct tipc_sock *tsk = tipc_sk(sk);
794 __poll_t revents = 0;
795
796 sock_poll_wait(file, sock, wait);
797 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
798
799 if (sk->sk_shutdown & RCV_SHUTDOWN)
800 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
801 if (sk->sk_shutdown == SHUTDOWN_MASK)
802 revents |= EPOLLHUP;
803
804 switch (sk->sk_state) {
805 case TIPC_ESTABLISHED:
806 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
807 revents |= EPOLLOUT;
808 fallthrough;
809 case TIPC_LISTEN:
810 case TIPC_CONNECTING:
811 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
812 revents |= EPOLLIN | EPOLLRDNORM;
813 break;
814 case TIPC_OPEN:
815 if (tsk->group_is_open && !tsk->cong_link_cnt)
816 revents |= EPOLLOUT;
817 if (!tipc_sk_type_connectionless(sk))
818 break;
819 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
820 break;
821 revents |= EPOLLIN | EPOLLRDNORM;
822 break;
823 case TIPC_DISCONNECTING:
824 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
825 break;
826 }
827 return revents;
828}
829
830/**
831 * tipc_sendmcast - send multicast message
832 * @sock: socket structure
833 * @ua: destination address struct
834 * @msg: message to send
835 * @dlen: length of data to send
836 * @timeout: timeout to wait for wakeup
837 *
838 * Called from function tipc_sendmsg(), which has done all sanity checks
839 * Return: the number of bytes sent on success, or errno
840 */
841static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua,
842 struct msghdr *msg, size_t dlen, long timeout)
843{
844 struct sock *sk = sock->sk;
845 struct tipc_sock *tsk = tipc_sk(sk);
846 struct tipc_msg *hdr = &tsk->phdr;
847 struct net *net = sock_net(sk);
848 int mtu = tipc_bcast_get_mtu(net);
849 struct sk_buff_head pkts;
850 struct tipc_nlist dsts;
851 int rc;
852
853 if (tsk->group)
854 return -EACCES;
855
856 /* Block or return if any destination link is congested */
857 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
858 if (unlikely(rc))
859 return rc;
860
861 /* Lookup destination nodes */
862 tipc_nlist_init(&dsts, tipc_own_addr(net));
863 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
864 if (!dsts.local && !dsts.remote)
865 return -EHOSTUNREACH;
866
867 /* Build message header */
868 msg_set_type(hdr, TIPC_MCAST_MSG);
869 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
870 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
871 msg_set_destport(hdr, 0);
872 msg_set_destnode(hdr, 0);
873 msg_set_nametype(hdr, ua->sr.type);
874 msg_set_namelower(hdr, ua->sr.lower);
875 msg_set_nameupper(hdr, ua->sr.upper);
876
877 /* Build message as chain of buffers */
878 __skb_queue_head_init(&pkts);
879 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
880
881 /* Send message if build was successful */
882 if (unlikely(rc == dlen)) {
883 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
884 TIPC_DUMP_SK_SNDQ, " ");
885 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
886 &tsk->cong_link_cnt);
887 }
888
889 tipc_nlist_purge(&dsts);
890
891 return rc ? rc : dlen;
892}
893
894/**
895 * tipc_send_group_msg - send a message to a member in the group
896 * @net: network namespace
897 * @tsk: tipc socket
898 * @m: message to send
899 * @mb: group member
900 * @dnode: destination node
901 * @dport: destination port
902 * @dlen: total length of message data
903 */
904static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
905 struct msghdr *m, struct tipc_member *mb,
906 u32 dnode, u32 dport, int dlen)
907{
908 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
909 struct tipc_mc_method *method = &tsk->mc_method;
910 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
911 struct tipc_msg *hdr = &tsk->phdr;
912 struct sk_buff_head pkts;
913 int mtu, rc;
914
915 /* Complete message header */
916 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
917 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
918 msg_set_destport(hdr, dport);
919 msg_set_destnode(hdr, dnode);
920 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
921
922 /* Build message as chain of buffers */
923 __skb_queue_head_init(&pkts);
924 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
925 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
926 if (unlikely(rc != dlen))
927 return rc;
928
929 /* Send message */
930 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
931 if (unlikely(rc == -ELINKCONG)) {
932 tipc_dest_push(&tsk->cong_links, dnode, 0);
933 tsk->cong_link_cnt++;
934 }
935
936 /* Update send window */
937 tipc_group_update_member(mb, blks);
938
939 /* A broadcast sent within next EXPIRE period must follow same path */
940 method->rcast = true;
941 method->mandatory = true;
942 return dlen;
943}
944
945/**
946 * tipc_send_group_unicast - send message to a member in the group
947 * @sock: socket structure
948 * @m: message to send
949 * @dlen: total length of message data
950 * @timeout: timeout to wait for wakeup
951 *
952 * Called from function tipc_sendmsg(), which has done all sanity checks
953 * Return: the number of bytes sent on success, or errno
954 */
955static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
956 int dlen, long timeout)
957{
958 struct sock *sk = sock->sk;
959 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
960 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
961 struct tipc_sock *tsk = tipc_sk(sk);
962 struct net *net = sock_net(sk);
963 struct tipc_member *mb = NULL;
964 u32 node, port;
965 int rc;
966
967 node = ua->sk.node;
968 port = ua->sk.ref;
969 if (!port && !node)
970 return -EHOSTUNREACH;
971
972 /* Block or return if destination link or member is congested */
973 rc = tipc_wait_for_cond(sock, &timeout,
974 !tipc_dest_find(&tsk->cong_links, node, 0) &&
975 tsk->group &&
976 !tipc_group_cong(tsk->group, node, port, blks,
977 &mb));
978 if (unlikely(rc))
979 return rc;
980
981 if (unlikely(!mb))
982 return -EHOSTUNREACH;
983
984 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
985
986 return rc ? rc : dlen;
987}
988
989/**
990 * tipc_send_group_anycast - send message to any member with given identity
991 * @sock: socket structure
992 * @m: message to send
993 * @dlen: total length of message data
994 * @timeout: timeout to wait for wakeup
995 *
996 * Called from function tipc_sendmsg(), which has done all sanity checks
997 * Return: the number of bytes sent on success, or errno
998 */
999static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1000 int dlen, long timeout)
1001{
1002 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1003 struct sock *sk = sock->sk;
1004 struct tipc_sock *tsk = tipc_sk(sk);
1005 struct list_head *cong_links = &tsk->cong_links;
1006 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1007 struct tipc_msg *hdr = &tsk->phdr;
1008 struct tipc_member *first = NULL;
1009 struct tipc_member *mbr = NULL;
1010 struct net *net = sock_net(sk);
1011 u32 node, port, exclude;
1012 LIST_HEAD(dsts);
1013 int lookups = 0;
1014 int dstcnt, rc;
1015 bool cong;
1016
1017 ua->sa.type = msg_nametype(hdr);
1018 ua->scope = msg_lookup_scope(hdr);
1019
1020 while (++lookups < 4) {
1021 exclude = tipc_group_exclude(tsk->group);
1022
1023 first = NULL;
1024
1025 /* Look for a non-congested destination member, if any */
1026 while (1) {
1027 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1028 exclude, false))
1029 return -EHOSTUNREACH;
1030 tipc_dest_pop(&dsts, &node, &port);
1031 cong = tipc_group_cong(tsk->group, node, port, blks,
1032 &mbr);
1033 if (!cong)
1034 break;
1035 if (mbr == first)
1036 break;
1037 if (!first)
1038 first = mbr;
1039 }
1040
1041 /* Start over if destination was not in member list */
1042 if (unlikely(!mbr))
1043 continue;
1044
1045 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1046 break;
1047
1048 /* Block or return if destination link or member is congested */
1049 rc = tipc_wait_for_cond(sock, &timeout,
1050 !tipc_dest_find(cong_links, node, 0) &&
1051 tsk->group &&
1052 !tipc_group_cong(tsk->group, node, port,
1053 blks, &mbr));
1054 if (unlikely(rc))
1055 return rc;
1056
1057 /* Send, unless destination disappeared while waiting */
1058 if (likely(mbr))
1059 break;
1060 }
1061
1062 if (unlikely(lookups >= 4))
1063 return -EHOSTUNREACH;
1064
1065 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1066
1067 return rc ? rc : dlen;
1068}
1069
1070/**
1071 * tipc_send_group_bcast - send message to all members in communication group
1072 * @sock: socket structure
1073 * @m: message to send
1074 * @dlen: total length of message data
1075 * @timeout: timeout to wait for wakeup
1076 *
1077 * Called from function tipc_sendmsg(), which has done all sanity checks
1078 * Return: the number of bytes sent on success, or errno
1079 */
1080static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1081 int dlen, long timeout)
1082{
1083 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1084 struct sock *sk = sock->sk;
1085 struct net *net = sock_net(sk);
1086 struct tipc_sock *tsk = tipc_sk(sk);
1087 struct tipc_nlist *dsts;
1088 struct tipc_mc_method *method = &tsk->mc_method;
1089 bool ack = method->mandatory && method->rcast;
1090 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1091 struct tipc_msg *hdr = &tsk->phdr;
1092 int mtu = tipc_bcast_get_mtu(net);
1093 struct sk_buff_head pkts;
1094 int rc = -EHOSTUNREACH;
1095
1096 /* Block or return if any destination link or member is congested */
1097 rc = tipc_wait_for_cond(sock, &timeout,
1098 !tsk->cong_link_cnt && tsk->group &&
1099 !tipc_group_bc_cong(tsk->group, blks));
1100 if (unlikely(rc))
1101 return rc;
1102
1103 dsts = tipc_group_dests(tsk->group);
1104 if (!dsts->local && !dsts->remote)
1105 return -EHOSTUNREACH;
1106
1107 /* Complete message header */
1108 if (ua) {
1109 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1110 msg_set_nameinst(hdr, ua->sa.instance);
1111 } else {
1112 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1113 msg_set_nameinst(hdr, 0);
1114 }
1115 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1116 msg_set_destport(hdr, 0);
1117 msg_set_destnode(hdr, 0);
1118 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1119
1120 /* Avoid getting stuck with repeated forced replicasts */
1121 msg_set_grp_bc_ack_req(hdr, ack);
1122
1123 /* Build message as chain of buffers */
1124 __skb_queue_head_init(&pkts);
1125 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1126 if (unlikely(rc != dlen))
1127 return rc;
1128
1129 /* Send message */
1130 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1131 if (unlikely(rc))
1132 return rc;
1133
1134 /* Update broadcast sequence number and send windows */
1135 tipc_group_update_bc_members(tsk->group, blks, ack);
1136
1137 /* Broadcast link is now free to choose method for next broadcast */
1138 method->mandatory = false;
1139 method->expires = jiffies;
1140
1141 return dlen;
1142}
1143
1144/**
1145 * tipc_send_group_mcast - send message to all members with given identity
1146 * @sock: socket structure
1147 * @m: message to send
1148 * @dlen: total length of message data
1149 * @timeout: timeout to wait for wakeup
1150 *
1151 * Called from function tipc_sendmsg(), which has done all sanity checks
1152 * Return: the number of bytes sent on success, or errno
1153 */
1154static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1155 int dlen, long timeout)
1156{
1157 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1158 struct sock *sk = sock->sk;
1159 struct tipc_sock *tsk = tipc_sk(sk);
1160 struct tipc_group *grp = tsk->group;
1161 struct tipc_msg *hdr = &tsk->phdr;
1162 struct net *net = sock_net(sk);
1163 u32 dstcnt, exclude;
1164 LIST_HEAD(dsts);
1165
1166 ua->sa.type = msg_nametype(hdr);
1167 ua->scope = msg_lookup_scope(hdr);
1168 exclude = tipc_group_exclude(grp);
1169
1170 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1171 return -EHOSTUNREACH;
1172
1173 if (dstcnt == 1) {
1174 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1175 return tipc_send_group_unicast(sock, m, dlen, timeout);
1176 }
1177
1178 tipc_dest_list_purge(&dsts);
1179 return tipc_send_group_bcast(sock, m, dlen, timeout);
1180}
1181
1182/**
1183 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1184 * @net: the associated network namespace
1185 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1186 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1187 *
1188 * Multi-threaded: parallel calls with reference to same queues may occur
1189 */
1190void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1191 struct sk_buff_head *inputq)
1192{
1193 u32 self = tipc_own_addr(net);
1194 struct sk_buff *skb, *_skb;
1195 u32 portid, onode;
1196 struct sk_buff_head tmpq;
1197 struct list_head dports;
1198 struct tipc_msg *hdr;
1199 struct tipc_uaddr ua;
1200 int user, mtyp, hlen;
1201
1202 __skb_queue_head_init(&tmpq);
1203 INIT_LIST_HEAD(&dports);
1204 ua.addrtype = TIPC_SERVICE_RANGE;
1205
1206 /* tipc_skb_peek() increments the head skb's reference counter */
1207 skb = tipc_skb_peek(arrvq, &inputq->lock);
1208 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1209 hdr = buf_msg(skb);
1210 user = msg_user(hdr);
1211 mtyp = msg_type(hdr);
1212 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1213 onode = msg_orignode(hdr);
1214 ua.sr.type = msg_nametype(hdr);
1215 ua.sr.lower = msg_namelower(hdr);
1216 ua.sr.upper = msg_nameupper(hdr);
1217 if (onode == self)
1218 ua.scope = TIPC_ANY_SCOPE;
1219 else
1220 ua.scope = TIPC_CLUSTER_SCOPE;
1221
1222 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1223 spin_lock_bh(&inputq->lock);
1224 if (skb_peek(arrvq) == skb) {
1225 __skb_dequeue(arrvq);
1226 __skb_queue_tail(inputq, skb);
1227 }
1228 kfree_skb(skb);
1229 spin_unlock_bh(&inputq->lock);
1230 continue;
1231 }
1232
1233 /* Group messages require exact scope match */
1234 if (msg_in_group(hdr)) {
1235 ua.sr.lower = 0;
1236 ua.sr.upper = ~0;
1237 ua.scope = msg_lookup_scope(hdr);
1238 }
1239
1240 /* Create destination port list: */
1241 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1242
1243 /* Clone message per destination */
1244 while (tipc_dest_pop(&dports, NULL, &portid)) {
1245 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1246 if (_skb) {
1247 msg_set_destport(buf_msg(_skb), portid);
1248 __skb_queue_tail(&tmpq, _skb);
1249 continue;
1250 }
1251 pr_warn("Failed to clone mcast rcv buffer\n");
1252 }
1253 /* Append clones to inputq only if skb is still head of arrvq */
1254 spin_lock_bh(&inputq->lock);
1255 if (skb_peek(arrvq) == skb) {
1256 skb_queue_splice_tail_init(&tmpq, inputq);
1257 /* Decrement the skb's refcnt */
1258 kfree_skb(__skb_dequeue(arrvq));
1259 }
1260 spin_unlock_bh(&inputq->lock);
1261 __skb_queue_purge(&tmpq);
1262 kfree_skb(skb);
1263 }
1264 tipc_sk_rcv(net, inputq);
1265}
1266
1267/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1268 * when socket is in Nagle mode
1269 */
1270static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1271{
1272 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1273 struct sk_buff *skb = skb_peek_tail(txq);
1274 struct net *net = sock_net(&tsk->sk);
1275 u32 dnode = tsk_peer_node(tsk);
1276 int rc;
1277
1278 if (nagle_ack) {
1279 tsk->pkt_cnt += skb_queue_len(txq);
1280 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1281 tsk->oneway = 0;
1282 if (tsk->nagle_start < NAGLE_START_MAX)
1283 tsk->nagle_start *= 2;
1284 tsk->expect_ack = false;
1285 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1286 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1287 tsk->nagle_start);
1288 } else {
1289 tsk->nagle_start = NAGLE_START_INIT;
1290 if (skb) {
1291 msg_set_ack_required(buf_msg(skb));
1292 tsk->expect_ack = true;
1293 } else {
1294 tsk->expect_ack = false;
1295 }
1296 }
1297 tsk->msg_acc = 0;
1298 tsk->pkt_cnt = 0;
1299 }
1300
1301 if (!skb || tsk->cong_link_cnt)
1302 return;
1303
1304 /* Do not send SYN again after congestion */
1305 if (msg_is_syn(buf_msg(skb)))
1306 return;
1307
1308 if (tsk->msg_acc)
1309 tsk->pkt_cnt += skb_queue_len(txq);
1310 tsk->snt_unacked += tsk->snd_backlog;
1311 tsk->snd_backlog = 0;
1312 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1313 if (rc == -ELINKCONG)
1314 tsk->cong_link_cnt = 1;
1315}
1316
1317/**
1318 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1319 * @tsk: receiving socket
1320 * @skb: pointer to message buffer.
1321 * @inputq: buffer list containing the buffers
1322 * @xmitq: output message area
1323 */
1324static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1325 struct sk_buff_head *inputq,
1326 struct sk_buff_head *xmitq)
1327{
1328 struct tipc_msg *hdr = buf_msg(skb);
1329 u32 onode = tsk_own_node(tsk);
1330 struct sock *sk = &tsk->sk;
1331 int mtyp = msg_type(hdr);
1332 bool was_cong;
1333
1334 /* Ignore if connection cannot be validated: */
1335 if (!tsk_peer_msg(tsk, hdr)) {
1336 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1337 goto exit;
1338 }
1339
1340 if (unlikely(msg_errcode(hdr))) {
1341 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1342 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1343 tsk_peer_port(tsk));
1344 sk->sk_state_change(sk);
1345
1346 /* State change is ignored if socket already awake,
1347 * - convert msg to abort msg and add to inqueue
1348 */
1349 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1350 msg_set_type(hdr, TIPC_CONN_MSG);
1351 msg_set_size(hdr, BASIC_H_SIZE);
1352 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1353 __skb_queue_tail(inputq, skb);
1354 return;
1355 }
1356
1357 tsk->probe_unacked = false;
1358
1359 if (mtyp == CONN_PROBE) {
1360 msg_set_type(hdr, CONN_PROBE_REPLY);
1361 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1362 __skb_queue_tail(xmitq, skb);
1363 return;
1364 } else if (mtyp == CONN_ACK) {
1365 was_cong = tsk_conn_cong(tsk);
1366 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1367 tsk->snt_unacked -= msg_conn_ack(hdr);
1368 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1369 tsk->snd_win = msg_adv_win(hdr);
1370 if (was_cong && !tsk_conn_cong(tsk))
1371 sk->sk_write_space(sk);
1372 } else if (mtyp != CONN_PROBE_REPLY) {
1373 pr_warn("Received unknown CONN_PROTO msg\n");
1374 }
1375exit:
1376 kfree_skb(skb);
1377}
1378
1379/**
1380 * tipc_sendmsg - send message in connectionless manner
1381 * @sock: socket structure
1382 * @m: message to send
1383 * @dsz: amount of user data to be sent
1384 *
1385 * Message must have an destination specified explicitly.
1386 * Used for SOCK_RDM and SOCK_DGRAM messages,
1387 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1388 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1389 *
1390 * Return: the number of bytes sent on success, or errno otherwise
1391 */
1392static int tipc_sendmsg(struct socket *sock,
1393 struct msghdr *m, size_t dsz)
1394{
1395 struct sock *sk = sock->sk;
1396 int ret;
1397
1398 lock_sock(sk);
1399 ret = __tipc_sendmsg(sock, m, dsz);
1400 release_sock(sk);
1401
1402 return ret;
1403}
1404
1405static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1406{
1407 struct sock *sk = sock->sk;
1408 struct net *net = sock_net(sk);
1409 struct tipc_sock *tsk = tipc_sk(sk);
1410 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1411 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1412 struct list_head *clinks = &tsk->cong_links;
1413 bool syn = !tipc_sk_type_connectionless(sk);
1414 struct tipc_group *grp = tsk->group;
1415 struct tipc_msg *hdr = &tsk->phdr;
1416 struct tipc_socket_addr skaddr;
1417 struct sk_buff_head pkts;
1418 int atype, mtu, rc;
1419
1420 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1421 return -EMSGSIZE;
1422
1423 if (ua) {
1424 if (!tipc_uaddr_valid(ua, m->msg_namelen))
1425 return -EINVAL;
1426 atype = ua->addrtype;
1427 }
1428
1429 /* If socket belongs to a communication group follow other paths */
1430 if (grp) {
1431 if (!ua)
1432 return tipc_send_group_bcast(sock, m, dlen, timeout);
1433 if (atype == TIPC_SERVICE_ADDR)
1434 return tipc_send_group_anycast(sock, m, dlen, timeout);
1435 if (atype == TIPC_SOCKET_ADDR)
1436 return tipc_send_group_unicast(sock, m, dlen, timeout);
1437 if (atype == TIPC_SERVICE_RANGE)
1438 return tipc_send_group_mcast(sock, m, dlen, timeout);
1439 return -EINVAL;
1440 }
1441
1442 if (!ua) {
1443 ua = (struct tipc_uaddr *)&tsk->peer;
1444 if (!syn && ua->family != AF_TIPC)
1445 return -EDESTADDRREQ;
1446 atype = ua->addrtype;
1447 }
1448
1449 if (unlikely(syn)) {
1450 if (sk->sk_state == TIPC_LISTEN)
1451 return -EPIPE;
1452 if (sk->sk_state != TIPC_OPEN)
1453 return -EISCONN;
1454 if (tsk->published)
1455 return -EOPNOTSUPP;
1456 if (atype == TIPC_SERVICE_ADDR)
1457 tsk->conn_addrtype = atype;
1458 msg_set_syn(hdr, 1);
1459 }
1460
1461 memset(&skaddr, 0, sizeof(skaddr));
1462
1463 /* Determine destination */
1464 if (atype == TIPC_SERVICE_RANGE) {
1465 return tipc_sendmcast(sock, ua, m, dlen, timeout);
1466 } else if (atype == TIPC_SERVICE_ADDR) {
1467 skaddr.node = ua->lookup_node;
1468 ua->scope = tipc_node2scope(skaddr.node);
1469 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1470 return -EHOSTUNREACH;
1471 } else if (atype == TIPC_SOCKET_ADDR) {
1472 skaddr = ua->sk;
1473 } else {
1474 return -EINVAL;
1475 }
1476
1477 /* Block or return if destination link is congested */
1478 rc = tipc_wait_for_cond(sock, &timeout,
1479 !tipc_dest_find(clinks, skaddr.node, 0));
1480 if (unlikely(rc))
1481 return rc;
1482
1483 /* Finally build message header */
1484 msg_set_destnode(hdr, skaddr.node);
1485 msg_set_destport(hdr, skaddr.ref);
1486 if (atype == TIPC_SERVICE_ADDR) {
1487 msg_set_type(hdr, TIPC_NAMED_MSG);
1488 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1489 msg_set_nametype(hdr, ua->sa.type);
1490 msg_set_nameinst(hdr, ua->sa.instance);
1491 msg_set_lookup_scope(hdr, ua->scope);
1492 } else { /* TIPC_SOCKET_ADDR */
1493 msg_set_type(hdr, TIPC_DIRECT_MSG);
1494 msg_set_lookup_scope(hdr, 0);
1495 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1496 }
1497
1498 /* Add message body */
1499 __skb_queue_head_init(&pkts);
1500 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1501 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1502 if (unlikely(rc != dlen))
1503 return rc;
1504 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1505 __skb_queue_purge(&pkts);
1506 return -ENOMEM;
1507 }
1508
1509 /* Send message */
1510 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1511 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1512 if (unlikely(rc == -ELINKCONG)) {
1513 tipc_dest_push(clinks, skaddr.node, 0);
1514 tsk->cong_link_cnt++;
1515 rc = 0;
1516 }
1517
1518 if (unlikely(syn && !rc)) {
1519 tipc_set_sk_state(sk, TIPC_CONNECTING);
1520 if (dlen && timeout) {
1521 timeout = msecs_to_jiffies(timeout);
1522 tipc_wait_for_connect(sock, &timeout);
1523 }
1524 }
1525
1526 return rc ? rc : dlen;
1527}
1528
1529/**
1530 * tipc_sendstream - send stream-oriented data
1531 * @sock: socket structure
1532 * @m: data to send
1533 * @dsz: total length of data to be transmitted
1534 *
1535 * Used for SOCK_STREAM data.
1536 *
1537 * Return: the number of bytes sent on success (or partial success),
1538 * or errno if no data sent
1539 */
1540static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1541{
1542 struct sock *sk = sock->sk;
1543 int ret;
1544
1545 lock_sock(sk);
1546 ret = __tipc_sendstream(sock, m, dsz);
1547 release_sock(sk);
1548
1549 return ret;
1550}
1551
1552static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1553{
1554 struct sock *sk = sock->sk;
1555 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1556 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1557 struct sk_buff_head *txq = &sk->sk_write_queue;
1558 struct tipc_sock *tsk = tipc_sk(sk);
1559 struct tipc_msg *hdr = &tsk->phdr;
1560 struct net *net = sock_net(sk);
1561 struct sk_buff *skb;
1562 u32 dnode = tsk_peer_node(tsk);
1563 int maxnagle = tsk->maxnagle;
1564 int maxpkt = tsk->max_pkt;
1565 int send, sent = 0;
1566 int blocks, rc = 0;
1567
1568 if (unlikely(dlen > INT_MAX))
1569 return -EMSGSIZE;
1570
1571 /* Handle implicit connection setup */
1572 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1573 rc = __tipc_sendmsg(sock, m, dlen);
1574 if (dlen && dlen == rc) {
1575 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1576 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1577 }
1578 return rc;
1579 }
1580
1581 do {
1582 rc = tipc_wait_for_cond(sock, &timeout,
1583 (!tsk->cong_link_cnt &&
1584 !tsk_conn_cong(tsk) &&
1585 tipc_sk_connected(sk)));
1586 if (unlikely(rc))
1587 break;
1588 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1589 blocks = tsk->snd_backlog;
1590 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1591 send <= maxnagle) {
1592 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1593 if (unlikely(rc < 0))
1594 break;
1595 blocks += rc;
1596 tsk->msg_acc++;
1597 if (blocks <= 64 && tsk->expect_ack) {
1598 tsk->snd_backlog = blocks;
1599 sent += send;
1600 break;
1601 } else if (blocks > 64) {
1602 tsk->pkt_cnt += skb_queue_len(txq);
1603 } else {
1604 skb = skb_peek_tail(txq);
1605 if (skb) {
1606 msg_set_ack_required(buf_msg(skb));
1607 tsk->expect_ack = true;
1608 } else {
1609 tsk->expect_ack = false;
1610 }
1611 tsk->msg_acc = 0;
1612 tsk->pkt_cnt = 0;
1613 }
1614 } else {
1615 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1616 if (unlikely(rc != send))
1617 break;
1618 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1619 }
1620 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1621 TIPC_DUMP_SK_SNDQ, " ");
1622 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1623 if (unlikely(rc == -ELINKCONG)) {
1624 tsk->cong_link_cnt = 1;
1625 rc = 0;
1626 }
1627 if (likely(!rc)) {
1628 tsk->snt_unacked += blocks;
1629 tsk->snd_backlog = 0;
1630 sent += send;
1631 }
1632 } while (sent < dlen && !rc);
1633
1634 return sent ? sent : rc;
1635}
1636
1637/**
1638 * tipc_send_packet - send a connection-oriented message
1639 * @sock: socket structure
1640 * @m: message to send
1641 * @dsz: length of data to be transmitted
1642 *
1643 * Used for SOCK_SEQPACKET messages.
1644 *
1645 * Return: the number of bytes sent on success, or errno otherwise
1646 */
1647static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1648{
1649 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1650 return -EMSGSIZE;
1651
1652 return tipc_sendstream(sock, m, dsz);
1653}
1654
1655/* tipc_sk_finish_conn - complete the setup of a connection
1656 */
1657static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1658 u32 peer_node)
1659{
1660 struct sock *sk = &tsk->sk;
1661 struct net *net = sock_net(sk);
1662 struct tipc_msg *msg = &tsk->phdr;
1663
1664 msg_set_syn(msg, 0);
1665 msg_set_destnode(msg, peer_node);
1666 msg_set_destport(msg, peer_port);
1667 msg_set_type(msg, TIPC_CONN_MSG);
1668 msg_set_lookup_scope(msg, 0);
1669 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1670
1671 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1672 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1673 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1674 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1675 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1676 tsk_set_nagle(tsk);
1677 __skb_queue_purge(&sk->sk_write_queue);
1678 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1679 return;
1680
1681 /* Fall back to message based flow control */
1682 tsk->rcv_win = FLOWCTL_MSG_WIN;
1683 tsk->snd_win = FLOWCTL_MSG_WIN;
1684}
1685
1686/**
1687 * tipc_sk_set_orig_addr - capture sender's address for received message
1688 * @m: descriptor for message info
1689 * @skb: received message
1690 *
1691 * Note: Address is not captured if not requested by receiver.
1692 */
1693static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1694{
1695 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1696 struct tipc_msg *hdr = buf_msg(skb);
1697
1698 if (!srcaddr)
1699 return;
1700
1701 srcaddr->sock.family = AF_TIPC;
1702 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1703 srcaddr->sock.scope = 0;
1704 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1705 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1706 srcaddr->sock.addr.name.domain = 0;
1707 m->msg_namelen = sizeof(struct sockaddr_tipc);
1708
1709 if (!msg_in_group(hdr))
1710 return;
1711
1712 /* Group message users may also want to know sending member's id */
1713 srcaddr->member.family = AF_TIPC;
1714 srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1715 srcaddr->member.scope = 0;
1716 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1717 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1718 srcaddr->member.addr.name.domain = 0;
1719 m->msg_namelen = sizeof(*srcaddr);
1720}
1721
1722/**
1723 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1724 * @m: descriptor for message info
1725 * @skb: received message buffer
1726 * @tsk: TIPC port associated with message
1727 *
1728 * Note: Ancillary data is not captured if not requested by receiver.
1729 *
1730 * Return: 0 if successful, otherwise errno
1731 */
1732static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1733 struct tipc_sock *tsk)
1734{
1735 struct tipc_msg *hdr;
1736 u32 data[3] = {0,};
1737 bool has_addr;
1738 int dlen, rc;
1739
1740 if (likely(m->msg_controllen == 0))
1741 return 0;
1742
1743 hdr = buf_msg(skb);
1744 dlen = msg_data_sz(hdr);
1745
1746 /* Capture errored message object, if any */
1747 if (msg_errcode(hdr)) {
1748 if (skb_linearize(skb))
1749 return -ENOMEM;
1750 hdr = buf_msg(skb);
1751 data[0] = msg_errcode(hdr);
1752 data[1] = dlen;
1753 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1754 if (rc || !dlen)
1755 return rc;
1756 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1757 if (rc)
1758 return rc;
1759 }
1760
1761 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1762 switch (msg_type(hdr)) {
1763 case TIPC_NAMED_MSG:
1764 has_addr = true;
1765 data[0] = msg_nametype(hdr);
1766 data[1] = msg_namelower(hdr);
1767 data[2] = data[1];
1768 break;
1769 case TIPC_MCAST_MSG:
1770 has_addr = true;
1771 data[0] = msg_nametype(hdr);
1772 data[1] = msg_namelower(hdr);
1773 data[2] = msg_nameupper(hdr);
1774 break;
1775 case TIPC_CONN_MSG:
1776 has_addr = !!tsk->conn_addrtype;
1777 data[0] = msg_nametype(&tsk->phdr);
1778 data[1] = msg_nameinst(&tsk->phdr);
1779 data[2] = data[1];
1780 break;
1781 default:
1782 has_addr = false;
1783 }
1784 if (!has_addr)
1785 return 0;
1786 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1787}
1788
1789static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1790{
1791 struct sock *sk = &tsk->sk;
1792 struct sk_buff *skb = NULL;
1793 struct tipc_msg *msg;
1794 u32 peer_port = tsk_peer_port(tsk);
1795 u32 dnode = tsk_peer_node(tsk);
1796
1797 if (!tipc_sk_connected(sk))
1798 return NULL;
1799 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1800 dnode, tsk_own_node(tsk), peer_port,
1801 tsk->portid, TIPC_OK);
1802 if (!skb)
1803 return NULL;
1804 msg = buf_msg(skb);
1805 msg_set_conn_ack(msg, tsk->rcv_unacked);
1806 tsk->rcv_unacked = 0;
1807
1808 /* Adjust to and advertize the correct window limit */
1809 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1810 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1811 msg_set_adv_win(msg, tsk->rcv_win);
1812 }
1813 return skb;
1814}
1815
1816static void tipc_sk_send_ack(struct tipc_sock *tsk)
1817{
1818 struct sk_buff *skb;
1819
1820 skb = tipc_sk_build_ack(tsk);
1821 if (!skb)
1822 return;
1823
1824 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1825 msg_link_selector(buf_msg(skb)));
1826}
1827
1828static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1829{
1830 struct sock *sk = sock->sk;
1831 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1832 long timeo = *timeop;
1833 int err = sock_error(sk);
1834
1835 if (err)
1836 return err;
1837
1838 for (;;) {
1839 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1840 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1841 err = -ENOTCONN;
1842 break;
1843 }
1844 add_wait_queue(sk_sleep(sk), &wait);
1845 release_sock(sk);
1846 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1847 sched_annotate_sleep();
1848 lock_sock(sk);
1849 remove_wait_queue(sk_sleep(sk), &wait);
1850 }
1851 err = 0;
1852 if (!skb_queue_empty(&sk->sk_receive_queue))
1853 break;
1854 err = -EAGAIN;
1855 if (!timeo)
1856 break;
1857 err = sock_intr_errno(timeo);
1858 if (signal_pending(current))
1859 break;
1860
1861 err = sock_error(sk);
1862 if (err)
1863 break;
1864 }
1865 *timeop = timeo;
1866 return err;
1867}
1868
1869/**
1870 * tipc_recvmsg - receive packet-oriented message
1871 * @sock: network socket
1872 * @m: descriptor for message info
1873 * @buflen: length of user buffer area
1874 * @flags: receive flags
1875 *
1876 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1877 * If the complete message doesn't fit in user area, truncate it.
1878 *
1879 * Return: size of returned message data, errno otherwise
1880 */
1881static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1882 size_t buflen, int flags)
1883{
1884 struct sock *sk = sock->sk;
1885 bool connected = !tipc_sk_type_connectionless(sk);
1886 struct tipc_sock *tsk = tipc_sk(sk);
1887 int rc, err, hlen, dlen, copy;
1888 struct tipc_skb_cb *skb_cb;
1889 struct sk_buff_head xmitq;
1890 struct tipc_msg *hdr;
1891 struct sk_buff *skb;
1892 bool grp_evt;
1893 long timeout;
1894
1895 /* Catch invalid receive requests */
1896 if (unlikely(!buflen))
1897 return -EINVAL;
1898
1899 lock_sock(sk);
1900 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1901 rc = -ENOTCONN;
1902 goto exit;
1903 }
1904 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1905
1906 /* Step rcv queue to first msg with data or error; wait if necessary */
1907 do {
1908 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1909 if (unlikely(rc))
1910 goto exit;
1911 skb = skb_peek(&sk->sk_receive_queue);
1912 skb_cb = TIPC_SKB_CB(skb);
1913 hdr = buf_msg(skb);
1914 dlen = msg_data_sz(hdr);
1915 hlen = msg_hdr_sz(hdr);
1916 err = msg_errcode(hdr);
1917 grp_evt = msg_is_grp_evt(hdr);
1918 if (likely(dlen || err))
1919 break;
1920 tsk_advance_rx_queue(sk);
1921 } while (1);
1922
1923 /* Collect msg meta data, including error code and rejected data */
1924 tipc_sk_set_orig_addr(m, skb);
1925 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1926 if (unlikely(rc))
1927 goto exit;
1928 hdr = buf_msg(skb);
1929
1930 /* Capture data if non-error msg, otherwise just set return value */
1931 if (likely(!err)) {
1932 int offset = skb_cb->bytes_read;
1933
1934 copy = min_t(int, dlen - offset, buflen);
1935 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1936 if (unlikely(rc))
1937 goto exit;
1938 if (unlikely(offset + copy < dlen)) {
1939 if (flags & MSG_EOR) {
1940 if (!(flags & MSG_PEEK))
1941 skb_cb->bytes_read = offset + copy;
1942 } else {
1943 m->msg_flags |= MSG_TRUNC;
1944 skb_cb->bytes_read = 0;
1945 }
1946 } else {
1947 if (flags & MSG_EOR)
1948 m->msg_flags |= MSG_EOR;
1949 skb_cb->bytes_read = 0;
1950 }
1951 } else {
1952 copy = 0;
1953 rc = 0;
1954 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1955 rc = -ECONNRESET;
1956 goto exit;
1957 }
1958 }
1959
1960 /* Mark message as group event if applicable */
1961 if (unlikely(grp_evt)) {
1962 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1963 m->msg_flags |= MSG_EOR;
1964 m->msg_flags |= MSG_OOB;
1965 copy = 0;
1966 }
1967
1968 /* Caption of data or error code/rejected data was successful */
1969 if (unlikely(flags & MSG_PEEK))
1970 goto exit;
1971
1972 /* Send group flow control advertisement when applicable */
1973 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1974 __skb_queue_head_init(&xmitq);
1975 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1976 msg_orignode(hdr), msg_origport(hdr),
1977 &xmitq);
1978 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1979 }
1980
1981 if (skb_cb->bytes_read)
1982 goto exit;
1983
1984 tsk_advance_rx_queue(sk);
1985
1986 if (likely(!connected))
1987 goto exit;
1988
1989 /* Send connection flow control advertisement when applicable */
1990 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1991 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1992 tipc_sk_send_ack(tsk);
1993exit:
1994 release_sock(sk);
1995 return rc ? rc : copy;
1996}
1997
1998/**
1999 * tipc_recvstream - receive stream-oriented data
2000 * @sock: network socket
2001 * @m: descriptor for message info
2002 * @buflen: total size of user buffer area
2003 * @flags: receive flags
2004 *
2005 * Used for SOCK_STREAM messages only. If not enough data is available
2006 * will optionally wait for more; never truncates data.
2007 *
2008 * Return: size of returned message data, errno otherwise
2009 */
2010static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2011 size_t buflen, int flags)
2012{
2013 struct sock *sk = sock->sk;
2014 struct tipc_sock *tsk = tipc_sk(sk);
2015 struct sk_buff *skb;
2016 struct tipc_msg *hdr;
2017 struct tipc_skb_cb *skb_cb;
2018 bool peek = flags & MSG_PEEK;
2019 int offset, required, copy, copied = 0;
2020 int hlen, dlen, err, rc;
2021 long timeout;
2022
2023 /* Catch invalid receive attempts */
2024 if (unlikely(!buflen))
2025 return -EINVAL;
2026
2027 lock_sock(sk);
2028
2029 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2030 rc = -ENOTCONN;
2031 goto exit;
2032 }
2033 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2034 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2035
2036 do {
2037 /* Look at first msg in receive queue; wait if necessary */
2038 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2039 if (unlikely(rc))
2040 break;
2041 skb = skb_peek(&sk->sk_receive_queue);
2042 skb_cb = TIPC_SKB_CB(skb);
2043 hdr = buf_msg(skb);
2044 dlen = msg_data_sz(hdr);
2045 hlen = msg_hdr_sz(hdr);
2046 err = msg_errcode(hdr);
2047
2048 /* Discard any empty non-errored (SYN-) message */
2049 if (unlikely(!dlen && !err)) {
2050 tsk_advance_rx_queue(sk);
2051 continue;
2052 }
2053
2054 /* Collect msg meta data, incl. error code and rejected data */
2055 if (!copied) {
2056 tipc_sk_set_orig_addr(m, skb);
2057 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2058 if (rc)
2059 break;
2060 hdr = buf_msg(skb);
2061 }
2062
2063 /* Copy data if msg ok, otherwise return error/partial data */
2064 if (likely(!err)) {
2065 offset = skb_cb->bytes_read;
2066 copy = min_t(int, dlen - offset, buflen - copied);
2067 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2068 if (unlikely(rc))
2069 break;
2070 copied += copy;
2071 offset += copy;
2072 if (unlikely(offset < dlen)) {
2073 if (!peek)
2074 skb_cb->bytes_read = offset;
2075 break;
2076 }
2077 } else {
2078 rc = 0;
2079 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2080 rc = -ECONNRESET;
2081 if (copied || rc)
2082 break;
2083 }
2084
2085 if (unlikely(peek))
2086 break;
2087
2088 tsk_advance_rx_queue(sk);
2089
2090 /* Send connection flow control advertisement when applicable */
2091 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2092 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2093 tipc_sk_send_ack(tsk);
2094
2095 /* Exit if all requested data or FIN/error received */
2096 if (copied == buflen || err)
2097 break;
2098
2099 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2100exit:
2101 release_sock(sk);
2102 return copied ? copied : rc;
2103}
2104
2105/**
2106 * tipc_write_space - wake up thread if port congestion is released
2107 * @sk: socket
2108 */
2109static void tipc_write_space(struct sock *sk)
2110{
2111 struct socket_wq *wq;
2112
2113 rcu_read_lock();
2114 wq = rcu_dereference(sk->sk_wq);
2115 if (skwq_has_sleeper(wq))
2116 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2117 EPOLLWRNORM | EPOLLWRBAND);
2118 rcu_read_unlock();
2119}
2120
2121/**
2122 * tipc_data_ready - wake up threads to indicate messages have been received
2123 * @sk: socket
2124 */
2125static void tipc_data_ready(struct sock *sk)
2126{
2127 struct socket_wq *wq;
2128
2129 trace_sk_data_ready(sk);
2130
2131 rcu_read_lock();
2132 wq = rcu_dereference(sk->sk_wq);
2133 if (skwq_has_sleeper(wq))
2134 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2135 EPOLLRDNORM | EPOLLRDBAND);
2136 rcu_read_unlock();
2137}
2138
2139static void tipc_sock_destruct(struct sock *sk)
2140{
2141 __skb_queue_purge(&sk->sk_receive_queue);
2142}
2143
2144static void tipc_sk_proto_rcv(struct sock *sk,
2145 struct sk_buff_head *inputq,
2146 struct sk_buff_head *xmitq)
2147{
2148 struct sk_buff *skb = __skb_dequeue(inputq);
2149 struct tipc_sock *tsk = tipc_sk(sk);
2150 struct tipc_msg *hdr = buf_msg(skb);
2151 struct tipc_group *grp = tsk->group;
2152 bool wakeup = false;
2153
2154 switch (msg_user(hdr)) {
2155 case CONN_MANAGER:
2156 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2157 return;
2158 case SOCK_WAKEUP:
2159 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2160 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2161 smp_wmb();
2162 tsk->cong_link_cnt--;
2163 wakeup = true;
2164 tipc_sk_push_backlog(tsk, false);
2165 break;
2166 case GROUP_PROTOCOL:
2167 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2168 break;
2169 case TOP_SRV:
2170 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2171 hdr, inputq, xmitq);
2172 break;
2173 default:
2174 break;
2175 }
2176
2177 if (wakeup)
2178 sk->sk_write_space(sk);
2179
2180 kfree_skb(skb);
2181}
2182
2183/**
2184 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2185 * @tsk: TIPC socket
2186 * @skb: pointer to message buffer.
2187 * @xmitq: for Nagle ACK if any
2188 * Return: true if message should be added to receive queue, false otherwise
2189 */
2190static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2191 struct sk_buff_head *xmitq)
2192{
2193 struct sock *sk = &tsk->sk;
2194 struct net *net = sock_net(sk);
2195 struct tipc_msg *hdr = buf_msg(skb);
2196 bool con_msg = msg_connected(hdr);
2197 u32 pport = tsk_peer_port(tsk);
2198 u32 pnode = tsk_peer_node(tsk);
2199 u32 oport = msg_origport(hdr);
2200 u32 onode = msg_orignode(hdr);
2201 int err = msg_errcode(hdr);
2202 unsigned long delay;
2203
2204 if (unlikely(msg_mcast(hdr)))
2205 return false;
2206 tsk->oneway = 0;
2207
2208 switch (sk->sk_state) {
2209 case TIPC_CONNECTING:
2210 /* Setup ACK */
2211 if (likely(con_msg)) {
2212 if (err)
2213 break;
2214 tipc_sk_finish_conn(tsk, oport, onode);
2215 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2216 /* ACK+ message with data is added to receive queue */
2217 if (msg_data_sz(hdr))
2218 return true;
2219 /* Empty ACK-, - wake up sleeping connect() and drop */
2220 sk->sk_state_change(sk);
2221 msg_set_dest_droppable(hdr, 1);
2222 return false;
2223 }
2224 /* Ignore connectionless message if not from listening socket */
2225 if (oport != pport || onode != pnode)
2226 return false;
2227
2228 /* Rejected SYN */
2229 if (err != TIPC_ERR_OVERLOAD)
2230 break;
2231
2232 /* Prepare for new setup attempt if we have a SYN clone */
2233 if (skb_queue_empty(&sk->sk_write_queue))
2234 break;
2235 get_random_bytes(&delay, 2);
2236 delay %= (tsk->conn_timeout / 4);
2237 delay = msecs_to_jiffies(delay + 100);
2238 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2239 return false;
2240 case TIPC_OPEN:
2241 case TIPC_DISCONNECTING:
2242 return false;
2243 case TIPC_LISTEN:
2244 /* Accept only SYN message */
2245 if (!msg_is_syn(hdr) &&
2246 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2247 return false;
2248 if (!con_msg && !err)
2249 return true;
2250 return false;
2251 case TIPC_ESTABLISHED:
2252 if (!skb_queue_empty(&sk->sk_write_queue))
2253 tipc_sk_push_backlog(tsk, false);
2254 /* Accept only connection-based messages sent by peer */
2255 if (likely(con_msg && !err && pport == oport &&
2256 pnode == onode)) {
2257 if (msg_ack_required(hdr)) {
2258 struct sk_buff *skb;
2259
2260 skb = tipc_sk_build_ack(tsk);
2261 if (skb) {
2262 msg_set_nagle_ack(buf_msg(skb));
2263 __skb_queue_tail(xmitq, skb);
2264 }
2265 }
2266 return true;
2267 }
2268 if (!tsk_peer_msg(tsk, hdr))
2269 return false;
2270 if (!err)
2271 return true;
2272 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2273 tipc_node_remove_conn(net, pnode, tsk->portid);
2274 sk->sk_state_change(sk);
2275 return true;
2276 default:
2277 pr_err("Unknown sk_state %u\n", sk->sk_state);
2278 }
2279 /* Abort connection setup attempt */
2280 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2281 sk->sk_err = ECONNREFUSED;
2282 sk->sk_state_change(sk);
2283 return true;
2284}
2285
2286/**
2287 * rcvbuf_limit - get proper overload limit of socket receive queue
2288 * @sk: socket
2289 * @skb: message
2290 *
2291 * For connection oriented messages, irrespective of importance,
2292 * default queue limit is 2 MB.
2293 *
2294 * For connectionless messages, queue limits are based on message
2295 * importance as follows:
2296 *
2297 * TIPC_LOW_IMPORTANCE (2 MB)
2298 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2299 * TIPC_HIGH_IMPORTANCE (8 MB)
2300 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2301 *
2302 * Return: overload limit according to corresponding message importance
2303 */
2304static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2305{
2306 struct tipc_sock *tsk = tipc_sk(sk);
2307 struct tipc_msg *hdr = buf_msg(skb);
2308
2309 if (unlikely(msg_in_group(hdr)))
2310 return READ_ONCE(sk->sk_rcvbuf);
2311
2312 if (unlikely(!msg_connected(hdr)))
2313 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2314
2315 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2316 return READ_ONCE(sk->sk_rcvbuf);
2317
2318 return FLOWCTL_MSG_LIM;
2319}
2320
2321/**
2322 * tipc_sk_filter_rcv - validate incoming message
2323 * @sk: socket
2324 * @skb: pointer to message.
2325 * @xmitq: output message area (FIXME)
2326 *
2327 * Enqueues message on receive queue if acceptable; optionally handles
2328 * disconnect indication for a connected socket.
2329 *
2330 * Called with socket lock already taken
2331 */
2332static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2333 struct sk_buff_head *xmitq)
2334{
2335 bool sk_conn = !tipc_sk_type_connectionless(sk);
2336 struct tipc_sock *tsk = tipc_sk(sk);
2337 struct tipc_group *grp = tsk->group;
2338 struct tipc_msg *hdr = buf_msg(skb);
2339 struct net *net = sock_net(sk);
2340 struct sk_buff_head inputq;
2341 int mtyp = msg_type(hdr);
2342 int limit, err = TIPC_OK;
2343
2344 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2345 TIPC_SKB_CB(skb)->bytes_read = 0;
2346 __skb_queue_head_init(&inputq);
2347 __skb_queue_tail(&inputq, skb);
2348
2349 if (unlikely(!msg_isdata(hdr)))
2350 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2351
2352 if (unlikely(grp))
2353 tipc_group_filter_msg(grp, &inputq, xmitq);
2354
2355 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2356 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2357
2358 /* Validate and add to receive buffer if there is space */
2359 while ((skb = __skb_dequeue(&inputq))) {
2360 hdr = buf_msg(skb);
2361 limit = rcvbuf_limit(sk, skb);
2362 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2363 (!sk_conn && msg_connected(hdr)) ||
2364 (!grp && msg_in_group(hdr)))
2365 err = TIPC_ERR_NO_PORT;
2366 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2367 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2368 "err_overload2!");
2369 atomic_inc(&sk->sk_drops);
2370 err = TIPC_ERR_OVERLOAD;
2371 }
2372
2373 if (unlikely(err)) {
2374 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2375 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2376 "@filter_rcv!");
2377 __skb_queue_tail(xmitq, skb);
2378 }
2379 err = TIPC_OK;
2380 continue;
2381 }
2382 __skb_queue_tail(&sk->sk_receive_queue, skb);
2383 skb_set_owner_r(skb, sk);
2384 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2385 "rcvq >90% allocated!");
2386 sk->sk_data_ready(sk);
2387 }
2388}
2389
2390/**
2391 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2392 * @sk: socket
2393 * @skb: message
2394 *
2395 * Caller must hold socket lock
2396 */
2397static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2398{
2399 unsigned int before = sk_rmem_alloc_get(sk);
2400 struct sk_buff_head xmitq;
2401 unsigned int added;
2402
2403 __skb_queue_head_init(&xmitq);
2404
2405 tipc_sk_filter_rcv(sk, skb, &xmitq);
2406 added = sk_rmem_alloc_get(sk) - before;
2407 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2408
2409 /* Send pending response/rejected messages, if any */
2410 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2411 return 0;
2412}
2413
2414/**
2415 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2416 * inputq and try adding them to socket or backlog queue
2417 * @inputq: list of incoming buffers with potentially different destinations
2418 * @sk: socket where the buffers should be enqueued
2419 * @dport: port number for the socket
2420 * @xmitq: output queue
2421 *
2422 * Caller must hold socket lock
2423 */
2424static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2425 u32 dport, struct sk_buff_head *xmitq)
2426{
2427 unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2428 struct sk_buff *skb;
2429 unsigned int lim;
2430 atomic_t *dcnt;
2431 u32 onode;
2432
2433 while (skb_queue_len(inputq)) {
2434 if (unlikely(time_after_eq(jiffies, time_limit)))
2435 return;
2436
2437 skb = tipc_skb_dequeue(inputq, dport);
2438 if (unlikely(!skb))
2439 return;
2440
2441 /* Add message directly to receive queue if possible */
2442 if (!sock_owned_by_user(sk)) {
2443 tipc_sk_filter_rcv(sk, skb, xmitq);
2444 continue;
2445 }
2446
2447 /* Try backlog, compensating for double-counted bytes */
2448 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2449 if (!sk->sk_backlog.len)
2450 atomic_set(dcnt, 0);
2451 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2452 if (likely(!sk_add_backlog(sk, skb, lim))) {
2453 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2454 "bklg & rcvq >90% allocated!");
2455 continue;
2456 }
2457
2458 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2459 /* Overload => reject message back to sender */
2460 onode = tipc_own_addr(sock_net(sk));
2461 atomic_inc(&sk->sk_drops);
2462 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2463 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2464 "@sk_enqueue!");
2465 __skb_queue_tail(xmitq, skb);
2466 }
2467 break;
2468 }
2469}
2470
2471/**
2472 * tipc_sk_rcv - handle a chain of incoming buffers
2473 * @net: the associated network namespace
2474 * @inputq: buffer list containing the buffers
2475 * Consumes all buffers in list until inputq is empty
2476 * Note: may be called in multiple threads referring to the same queue
2477 */
2478void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2479{
2480 struct sk_buff_head xmitq;
2481 u32 dnode, dport = 0;
2482 int err;
2483 struct tipc_sock *tsk;
2484 struct sock *sk;
2485 struct sk_buff *skb;
2486
2487 __skb_queue_head_init(&xmitq);
2488 while (skb_queue_len(inputq)) {
2489 dport = tipc_skb_peek_port(inputq, dport);
2490 tsk = tipc_sk_lookup(net, dport);
2491
2492 if (likely(tsk)) {
2493 sk = &tsk->sk;
2494 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2495 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2496 spin_unlock_bh(&sk->sk_lock.slock);
2497 }
2498 /* Send pending response/rejected messages, if any */
2499 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2500 sock_put(sk);
2501 continue;
2502 }
2503 /* No destination socket => dequeue skb if still there */
2504 skb = tipc_skb_dequeue(inputq, dport);
2505 if (!skb)
2506 return;
2507
2508 /* Try secondary lookup if unresolved named message */
2509 err = TIPC_ERR_NO_PORT;
2510 if (tipc_msg_lookup_dest(net, skb, &err))
2511 goto xmit;
2512
2513 /* Prepare for message rejection */
2514 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2515 continue;
2516
2517 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2518xmit:
2519 dnode = msg_destnode(buf_msg(skb));
2520 tipc_node_xmit_skb(net, skb, dnode, dport);
2521 }
2522}
2523
2524static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2525{
2526 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2527 struct sock *sk = sock->sk;
2528 int done;
2529
2530 do {
2531 int err = sock_error(sk);
2532 if (err)
2533 return err;
2534 if (!*timeo_p)
2535 return -ETIMEDOUT;
2536 if (signal_pending(current))
2537 return sock_intr_errno(*timeo_p);
2538 if (sk->sk_state == TIPC_DISCONNECTING)
2539 break;
2540
2541 add_wait_queue(sk_sleep(sk), &wait);
2542 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2543 &wait);
2544 remove_wait_queue(sk_sleep(sk), &wait);
2545 } while (!done);
2546 return 0;
2547}
2548
2549static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2550{
2551 if (addr->family != AF_TIPC)
2552 return false;
2553 if (addr->addrtype == TIPC_SERVICE_RANGE)
2554 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2555 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2556 addr->addrtype == TIPC_SOCKET_ADDR);
2557}
2558
2559/**
2560 * tipc_connect - establish a connection to another TIPC port
2561 * @sock: socket structure
2562 * @dest: socket address for destination port
2563 * @destlen: size of socket address data structure
2564 * @flags: file-related flags associated with socket
2565 *
2566 * Return: 0 on success, errno otherwise
2567 */
2568static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2569 int destlen, int flags)
2570{
2571 struct sock *sk = sock->sk;
2572 struct tipc_sock *tsk = tipc_sk(sk);
2573 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2574 struct msghdr m = {NULL,};
2575 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2576 int previous;
2577 int res = 0;
2578
2579 if (destlen != sizeof(struct sockaddr_tipc))
2580 return -EINVAL;
2581
2582 lock_sock(sk);
2583
2584 if (tsk->group) {
2585 res = -EINVAL;
2586 goto exit;
2587 }
2588
2589 if (dst->family == AF_UNSPEC) {
2590 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2591 if (!tipc_sk_type_connectionless(sk))
2592 res = -EINVAL;
2593 goto exit;
2594 }
2595 if (!tipc_sockaddr_is_sane(dst)) {
2596 res = -EINVAL;
2597 goto exit;
2598 }
2599 /* DGRAM/RDM connect(), just save the destaddr */
2600 if (tipc_sk_type_connectionless(sk)) {
2601 memcpy(&tsk->peer, dest, destlen);
2602 goto exit;
2603 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2604 res = -EINVAL;
2605 goto exit;
2606 }
2607
2608 previous = sk->sk_state;
2609
2610 switch (sk->sk_state) {
2611 case TIPC_OPEN:
2612 /* Send a 'SYN-' to destination */
2613 m.msg_name = dest;
2614 m.msg_namelen = destlen;
2615 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2616
2617 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2618 * indicate send_msg() is never blocked.
2619 */
2620 if (!timeout)
2621 m.msg_flags = MSG_DONTWAIT;
2622
2623 res = __tipc_sendmsg(sock, &m, 0);
2624 if ((res < 0) && (res != -EWOULDBLOCK))
2625 goto exit;
2626
2627 /* Just entered TIPC_CONNECTING state; the only
2628 * difference is that return value in non-blocking
2629 * case is EINPROGRESS, rather than EALREADY.
2630 */
2631 res = -EINPROGRESS;
2632 fallthrough;
2633 case TIPC_CONNECTING:
2634 if (!timeout) {
2635 if (previous == TIPC_CONNECTING)
2636 res = -EALREADY;
2637 goto exit;
2638 }
2639 timeout = msecs_to_jiffies(timeout);
2640 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2641 res = tipc_wait_for_connect(sock, &timeout);
2642 break;
2643 case TIPC_ESTABLISHED:
2644 res = -EISCONN;
2645 break;
2646 default:
2647 res = -EINVAL;
2648 }
2649
2650exit:
2651 release_sock(sk);
2652 return res;
2653}
2654
2655/**
2656 * tipc_listen - allow socket to listen for incoming connections
2657 * @sock: socket structure
2658 * @len: (unused)
2659 *
2660 * Return: 0 on success, errno otherwise
2661 */
2662static int tipc_listen(struct socket *sock, int len)
2663{
2664 struct sock *sk = sock->sk;
2665 int res;
2666
2667 lock_sock(sk);
2668 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2669 release_sock(sk);
2670
2671 return res;
2672}
2673
2674static int tipc_wait_for_accept(struct socket *sock, long timeo)
2675{
2676 struct sock *sk = sock->sk;
2677 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2678 int err;
2679
2680 /* True wake-one mechanism for incoming connections: only
2681 * one process gets woken up, not the 'whole herd'.
2682 * Since we do not 'race & poll' for established sockets
2683 * anymore, the common case will execute the loop only once.
2684 */
2685 for (;;) {
2686 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2687 add_wait_queue(sk_sleep(sk), &wait);
2688 release_sock(sk);
2689 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2690 lock_sock(sk);
2691 remove_wait_queue(sk_sleep(sk), &wait);
2692 }
2693 err = 0;
2694 if (!skb_queue_empty(&sk->sk_receive_queue))
2695 break;
2696 err = -EAGAIN;
2697 if (!timeo)
2698 break;
2699 err = sock_intr_errno(timeo);
2700 if (signal_pending(current))
2701 break;
2702 }
2703 return err;
2704}
2705
2706/**
2707 * tipc_accept - wait for connection request
2708 * @sock: listening socket
2709 * @new_sock: new socket that is to be connected
2710 * @arg: arguments for accept
2711 *
2712 * Return: 0 on success, errno otherwise
2713 */
2714static int tipc_accept(struct socket *sock, struct socket *new_sock,
2715 struct proto_accept_arg *arg)
2716{
2717 struct sock *new_sk, *sk = sock->sk;
2718 struct tipc_sock *new_tsock;
2719 struct msghdr m = {NULL,};
2720 struct tipc_msg *msg;
2721 struct sk_buff *buf;
2722 long timeo;
2723 int res;
2724
2725 lock_sock(sk);
2726
2727 if (sk->sk_state != TIPC_LISTEN) {
2728 res = -EINVAL;
2729 goto exit;
2730 }
2731 timeo = sock_rcvtimeo(sk, arg->flags & O_NONBLOCK);
2732 res = tipc_wait_for_accept(sock, timeo);
2733 if (res)
2734 goto exit;
2735
2736 buf = skb_peek(&sk->sk_receive_queue);
2737
2738 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, arg->kern);
2739 if (res)
2740 goto exit;
2741 security_sk_clone(sock->sk, new_sock->sk);
2742
2743 new_sk = new_sock->sk;
2744 new_tsock = tipc_sk(new_sk);
2745 msg = buf_msg(buf);
2746
2747 /* we lock on new_sk; but lockdep sees the lock on sk */
2748 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2749
2750 /*
2751 * Reject any stray messages received by new socket
2752 * before the socket lock was taken (very, very unlikely)
2753 */
2754 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2755
2756 /* Connect new socket to it's peer */
2757 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2758
2759 tsk_set_importance(new_sk, msg_importance(msg));
2760 if (msg_named(msg)) {
2761 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2762 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2763 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2764 }
2765
2766 /*
2767 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2768 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2769 */
2770 if (!msg_data_sz(msg)) {
2771 tsk_advance_rx_queue(sk);
2772 } else {
2773 __skb_dequeue(&sk->sk_receive_queue);
2774 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2775 skb_set_owner_r(buf, new_sk);
2776 }
2777 iov_iter_kvec(&m.msg_iter, ITER_SOURCE, NULL, 0, 0);
2778 __tipc_sendstream(new_sock, &m, 0);
2779 release_sock(new_sk);
2780exit:
2781 release_sock(sk);
2782 return res;
2783}
2784
2785/**
2786 * tipc_shutdown - shutdown socket connection
2787 * @sock: socket structure
2788 * @how: direction to close (must be SHUT_RDWR)
2789 *
2790 * Terminates connection (if necessary), then purges socket's receive queue.
2791 *
2792 * Return: 0 on success, errno otherwise
2793 */
2794static int tipc_shutdown(struct socket *sock, int how)
2795{
2796 struct sock *sk = sock->sk;
2797 int res;
2798
2799 if (how != SHUT_RDWR)
2800 return -EINVAL;
2801
2802 lock_sock(sk);
2803
2804 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2805 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2806 sk->sk_shutdown = SHUTDOWN_MASK;
2807
2808 if (sk->sk_state == TIPC_DISCONNECTING) {
2809 /* Discard any unreceived messages */
2810 __skb_queue_purge(&sk->sk_receive_queue);
2811
2812 res = 0;
2813 } else {
2814 res = -ENOTCONN;
2815 }
2816 /* Wake up anyone sleeping in poll. */
2817 sk->sk_state_change(sk);
2818
2819 release_sock(sk);
2820 return res;
2821}
2822
2823static void tipc_sk_check_probing_state(struct sock *sk,
2824 struct sk_buff_head *list)
2825{
2826 struct tipc_sock *tsk = tipc_sk(sk);
2827 u32 pnode = tsk_peer_node(tsk);
2828 u32 pport = tsk_peer_port(tsk);
2829 u32 self = tsk_own_node(tsk);
2830 u32 oport = tsk->portid;
2831 struct sk_buff *skb;
2832
2833 if (tsk->probe_unacked) {
2834 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2835 sk->sk_err = ECONNABORTED;
2836 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2837 sk->sk_state_change(sk);
2838 return;
2839 }
2840 /* Prepare new probe */
2841 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2842 pnode, self, pport, oport, TIPC_OK);
2843 if (skb)
2844 __skb_queue_tail(list, skb);
2845 tsk->probe_unacked = true;
2846 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2847}
2848
2849static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2850{
2851 struct tipc_sock *tsk = tipc_sk(sk);
2852
2853 /* Try again later if dest link is congested */
2854 if (tsk->cong_link_cnt) {
2855 sk_reset_timer(sk, &sk->sk_timer,
2856 jiffies + msecs_to_jiffies(100));
2857 return;
2858 }
2859 /* Prepare SYN for retransmit */
2860 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2861}
2862
2863static void tipc_sk_timeout(struct timer_list *t)
2864{
2865 struct sock *sk = from_timer(sk, t, sk_timer);
2866 struct tipc_sock *tsk = tipc_sk(sk);
2867 u32 pnode = tsk_peer_node(tsk);
2868 struct sk_buff_head list;
2869 int rc = 0;
2870
2871 __skb_queue_head_init(&list);
2872 bh_lock_sock(sk);
2873
2874 /* Try again later if socket is busy */
2875 if (sock_owned_by_user(sk)) {
2876 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2877 bh_unlock_sock(sk);
2878 sock_put(sk);
2879 return;
2880 }
2881
2882 if (sk->sk_state == TIPC_ESTABLISHED)
2883 tipc_sk_check_probing_state(sk, &list);
2884 else if (sk->sk_state == TIPC_CONNECTING)
2885 tipc_sk_retry_connect(sk, &list);
2886
2887 bh_unlock_sock(sk);
2888
2889 if (!skb_queue_empty(&list))
2890 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2891
2892 /* SYN messages may cause link congestion */
2893 if (rc == -ELINKCONG) {
2894 tipc_dest_push(&tsk->cong_links, pnode, 0);
2895 tsk->cong_link_cnt = 1;
2896 }
2897 sock_put(sk);
2898}
2899
2900static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2901{
2902 struct sock *sk = &tsk->sk;
2903 struct net *net = sock_net(sk);
2904 struct tipc_socket_addr skaddr;
2905 struct publication *p;
2906 u32 key;
2907
2908 if (tipc_sk_connected(sk))
2909 return -EINVAL;
2910 key = tsk->portid + tsk->pub_count + 1;
2911 if (key == tsk->portid)
2912 return -EADDRINUSE;
2913 skaddr.ref = tsk->portid;
2914 skaddr.node = tipc_own_addr(net);
2915 p = tipc_nametbl_publish(net, ua, &skaddr, key);
2916 if (unlikely(!p))
2917 return -EINVAL;
2918
2919 list_add(&p->binding_sock, &tsk->publications);
2920 tsk->pub_count++;
2921 tsk->published = true;
2922 return 0;
2923}
2924
2925static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2926{
2927 struct net *net = sock_net(&tsk->sk);
2928 struct publication *safe, *p;
2929 struct tipc_uaddr _ua;
2930 int rc = -EINVAL;
2931
2932 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2933 if (!ua) {
2934 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2935 p->sr.type, p->sr.lower, p->sr.upper);
2936 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2937 continue;
2938 }
2939 /* Unbind specific publication */
2940 if (p->scope != ua->scope)
2941 continue;
2942 if (p->sr.type != ua->sr.type)
2943 continue;
2944 if (p->sr.lower != ua->sr.lower)
2945 continue;
2946 if (p->sr.upper != ua->sr.upper)
2947 break;
2948 tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2949 rc = 0;
2950 break;
2951 }
2952 if (list_empty(&tsk->publications)) {
2953 tsk->published = 0;
2954 rc = 0;
2955 }
2956 return rc;
2957}
2958
2959/* tipc_sk_reinit: set non-zero address in all existing sockets
2960 * when we go from standalone to network mode.
2961 */
2962void tipc_sk_reinit(struct net *net)
2963{
2964 struct tipc_net *tn = net_generic(net, tipc_net_id);
2965 struct rhashtable_iter iter;
2966 struct tipc_sock *tsk;
2967 struct tipc_msg *msg;
2968
2969 rhashtable_walk_enter(&tn->sk_rht, &iter);
2970
2971 do {
2972 rhashtable_walk_start(&iter);
2973
2974 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2975 sock_hold(&tsk->sk);
2976 rhashtable_walk_stop(&iter);
2977 lock_sock(&tsk->sk);
2978 msg = &tsk->phdr;
2979 msg_set_prevnode(msg, tipc_own_addr(net));
2980 msg_set_orignode(msg, tipc_own_addr(net));
2981 release_sock(&tsk->sk);
2982 rhashtable_walk_start(&iter);
2983 sock_put(&tsk->sk);
2984 }
2985
2986 rhashtable_walk_stop(&iter);
2987 } while (tsk == ERR_PTR(-EAGAIN));
2988
2989 rhashtable_walk_exit(&iter);
2990}
2991
2992static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2993{
2994 struct tipc_net *tn = net_generic(net, tipc_net_id);
2995 struct tipc_sock *tsk;
2996
2997 rcu_read_lock();
2998 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2999 if (tsk)
3000 sock_hold(&tsk->sk);
3001 rcu_read_unlock();
3002
3003 return tsk;
3004}
3005
3006static int tipc_sk_insert(struct tipc_sock *tsk)
3007{
3008 struct sock *sk = &tsk->sk;
3009 struct net *net = sock_net(sk);
3010 struct tipc_net *tn = net_generic(net, tipc_net_id);
3011 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3012 u32 portid = get_random_u32_below(remaining) + TIPC_MIN_PORT;
3013
3014 while (remaining--) {
3015 portid++;
3016 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3017 portid = TIPC_MIN_PORT;
3018 tsk->portid = portid;
3019 sock_hold(&tsk->sk);
3020 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3021 tsk_rht_params))
3022 return 0;
3023 sock_put(&tsk->sk);
3024 }
3025
3026 return -1;
3027}
3028
3029static void tipc_sk_remove(struct tipc_sock *tsk)
3030{
3031 struct sock *sk = &tsk->sk;
3032 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3033
3034 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3035 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3036 __sock_put(sk);
3037 }
3038}
3039
3040static const struct rhashtable_params tsk_rht_params = {
3041 .nelem_hint = 192,
3042 .head_offset = offsetof(struct tipc_sock, node),
3043 .key_offset = offsetof(struct tipc_sock, portid),
3044 .key_len = sizeof(u32), /* portid */
3045 .max_size = 1048576,
3046 .min_size = 256,
3047 .automatic_shrinking = true,
3048};
3049
3050int tipc_sk_rht_init(struct net *net)
3051{
3052 struct tipc_net *tn = net_generic(net, tipc_net_id);
3053
3054 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3055}
3056
3057void tipc_sk_rht_destroy(struct net *net)
3058{
3059 struct tipc_net *tn = net_generic(net, tipc_net_id);
3060
3061 /* Wait for socket readers to complete */
3062 synchronize_net();
3063
3064 rhashtable_destroy(&tn->sk_rht);
3065}
3066
3067static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3068{
3069 struct net *net = sock_net(&tsk->sk);
3070 struct tipc_group *grp = tsk->group;
3071 struct tipc_msg *hdr = &tsk->phdr;
3072 struct tipc_uaddr ua;
3073 int rc;
3074
3075 if (mreq->type < TIPC_RESERVED_TYPES)
3076 return -EACCES;
3077 if (mreq->scope > TIPC_NODE_SCOPE)
3078 return -EINVAL;
3079 if (mreq->scope != TIPC_NODE_SCOPE)
3080 mreq->scope = TIPC_CLUSTER_SCOPE;
3081 if (grp)
3082 return -EACCES;
3083 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3084 if (!grp)
3085 return -ENOMEM;
3086 tsk->group = grp;
3087 msg_set_lookup_scope(hdr, mreq->scope);
3088 msg_set_nametype(hdr, mreq->type);
3089 msg_set_dest_droppable(hdr, true);
3090 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3091 mreq->type, mreq->instance, mreq->instance);
3092 tipc_nametbl_build_group(net, grp, &ua);
3093 rc = tipc_sk_publish(tsk, &ua);
3094 if (rc) {
3095 tipc_group_delete(net, grp);
3096 tsk->group = NULL;
3097 return rc;
3098 }
3099 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3100 tsk->mc_method.rcast = true;
3101 tsk->mc_method.mandatory = true;
3102 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3103 return rc;
3104}
3105
3106static int tipc_sk_leave(struct tipc_sock *tsk)
3107{
3108 struct net *net = sock_net(&tsk->sk);
3109 struct tipc_group *grp = tsk->group;
3110 struct tipc_uaddr ua;
3111 int scope;
3112
3113 if (!grp)
3114 return -EINVAL;
3115 ua.addrtype = TIPC_SERVICE_RANGE;
3116 tipc_group_self(grp, &ua.sr, &scope);
3117 ua.scope = scope;
3118 tipc_group_delete(net, grp);
3119 tsk->group = NULL;
3120 tipc_sk_withdraw(tsk, &ua);
3121 return 0;
3122}
3123
3124/**
3125 * tipc_setsockopt - set socket option
3126 * @sock: socket structure
3127 * @lvl: option level
3128 * @opt: option identifier
3129 * @ov: pointer to new option value
3130 * @ol: length of option value
3131 *
3132 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3133 * (to ease compatibility).
3134 *
3135 * Return: 0 on success, errno otherwise
3136 */
3137static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3138 sockptr_t ov, unsigned int ol)
3139{
3140 struct sock *sk = sock->sk;
3141 struct tipc_sock *tsk = tipc_sk(sk);
3142 struct tipc_group_req mreq;
3143 u32 value = 0;
3144 int res = 0;
3145
3146 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3147 return 0;
3148 if (lvl != SOL_TIPC)
3149 return -ENOPROTOOPT;
3150
3151 switch (opt) {
3152 case TIPC_IMPORTANCE:
3153 case TIPC_SRC_DROPPABLE:
3154 case TIPC_DEST_DROPPABLE:
3155 case TIPC_CONN_TIMEOUT:
3156 case TIPC_NODELAY:
3157 if (ol < sizeof(value))
3158 return -EINVAL;
3159 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3160 return -EFAULT;
3161 break;
3162 case TIPC_GROUP_JOIN:
3163 if (ol < sizeof(mreq))
3164 return -EINVAL;
3165 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3166 return -EFAULT;
3167 break;
3168 default:
3169 if (!sockptr_is_null(ov) || ol)
3170 return -EINVAL;
3171 }
3172
3173 lock_sock(sk);
3174
3175 switch (opt) {
3176 case TIPC_IMPORTANCE:
3177 res = tsk_set_importance(sk, value);
3178 break;
3179 case TIPC_SRC_DROPPABLE:
3180 if (sock->type != SOCK_STREAM)
3181 tsk_set_unreliable(tsk, value);
3182 else
3183 res = -ENOPROTOOPT;
3184 break;
3185 case TIPC_DEST_DROPPABLE:
3186 tsk_set_unreturnable(tsk, value);
3187 break;
3188 case TIPC_CONN_TIMEOUT:
3189 tipc_sk(sk)->conn_timeout = value;
3190 break;
3191 case TIPC_MCAST_BROADCAST:
3192 tsk->mc_method.rcast = false;
3193 tsk->mc_method.mandatory = true;
3194 break;
3195 case TIPC_MCAST_REPLICAST:
3196 tsk->mc_method.rcast = true;
3197 tsk->mc_method.mandatory = true;
3198 break;
3199 case TIPC_GROUP_JOIN:
3200 res = tipc_sk_join(tsk, &mreq);
3201 break;
3202 case TIPC_GROUP_LEAVE:
3203 res = tipc_sk_leave(tsk);
3204 break;
3205 case TIPC_NODELAY:
3206 tsk->nodelay = !!value;
3207 tsk_set_nagle(tsk);
3208 break;
3209 default:
3210 res = -EINVAL;
3211 }
3212
3213 release_sock(sk);
3214
3215 return res;
3216}
3217
3218/**
3219 * tipc_getsockopt - get socket option
3220 * @sock: socket structure
3221 * @lvl: option level
3222 * @opt: option identifier
3223 * @ov: receptacle for option value
3224 * @ol: receptacle for length of option value
3225 *
3226 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3227 * (to ease compatibility).
3228 *
3229 * Return: 0 on success, errno otherwise
3230 */
3231static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3232 char __user *ov, int __user *ol)
3233{
3234 struct sock *sk = sock->sk;
3235 struct tipc_sock *tsk = tipc_sk(sk);
3236 struct tipc_service_range seq;
3237 int len, scope;
3238 u32 value;
3239 int res;
3240
3241 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3242 return put_user(0, ol);
3243 if (lvl != SOL_TIPC)
3244 return -ENOPROTOOPT;
3245 res = get_user(len, ol);
3246 if (res)
3247 return res;
3248
3249 lock_sock(sk);
3250
3251 switch (opt) {
3252 case TIPC_IMPORTANCE:
3253 value = tsk_importance(tsk);
3254 break;
3255 case TIPC_SRC_DROPPABLE:
3256 value = tsk_unreliable(tsk);
3257 break;
3258 case TIPC_DEST_DROPPABLE:
3259 value = tsk_unreturnable(tsk);
3260 break;
3261 case TIPC_CONN_TIMEOUT:
3262 value = tsk->conn_timeout;
3263 /* no need to set "res", since already 0 at this point */
3264 break;
3265 case TIPC_NODE_RECVQ_DEPTH:
3266 value = 0; /* was tipc_queue_size, now obsolete */
3267 break;
3268 case TIPC_SOCK_RECVQ_DEPTH:
3269 value = skb_queue_len(&sk->sk_receive_queue);
3270 break;
3271 case TIPC_SOCK_RECVQ_USED:
3272 value = sk_rmem_alloc_get(sk);
3273 break;
3274 case TIPC_GROUP_JOIN:
3275 seq.type = 0;
3276 if (tsk->group)
3277 tipc_group_self(tsk->group, &seq, &scope);
3278 value = seq.type;
3279 break;
3280 default:
3281 res = -EINVAL;
3282 }
3283
3284 release_sock(sk);
3285
3286 if (res)
3287 return res; /* "get" failed */
3288
3289 if (len < sizeof(value))
3290 return -EINVAL;
3291
3292 if (copy_to_user(ov, &value, sizeof(value)))
3293 return -EFAULT;
3294
3295 return put_user(sizeof(value), ol);
3296}
3297
3298static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3299{
3300 struct net *net = sock_net(sock->sk);
3301 struct tipc_sioc_nodeid_req nr = {0};
3302 struct tipc_sioc_ln_req lnr;
3303 void __user *argp = (void __user *)arg;
3304
3305 switch (cmd) {
3306 case SIOCGETLINKNAME:
3307 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3308 return -EFAULT;
3309 if (!tipc_node_get_linkname(net,
3310 lnr.bearer_id & 0xffff, lnr.peer,
3311 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3312 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3313 return -EFAULT;
3314 return 0;
3315 }
3316 return -EADDRNOTAVAIL;
3317 case SIOCGETNODEID:
3318 if (copy_from_user(&nr, argp, sizeof(nr)))
3319 return -EFAULT;
3320 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3321 return -EADDRNOTAVAIL;
3322 if (copy_to_user(argp, &nr, sizeof(nr)))
3323 return -EFAULT;
3324 return 0;
3325 default:
3326 return -ENOIOCTLCMD;
3327 }
3328}
3329
3330static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3331{
3332 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3333 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3334 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3335
3336 tsk1->peer.family = AF_TIPC;
3337 tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3338 tsk1->peer.scope = TIPC_NODE_SCOPE;
3339 tsk1->peer.addr.id.ref = tsk2->portid;
3340 tsk1->peer.addr.id.node = onode;
3341 tsk2->peer.family = AF_TIPC;
3342 tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3343 tsk2->peer.scope = TIPC_NODE_SCOPE;
3344 tsk2->peer.addr.id.ref = tsk1->portid;
3345 tsk2->peer.addr.id.node = onode;
3346
3347 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3348 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3349 return 0;
3350}
3351
3352/* Protocol switches for the various types of TIPC sockets */
3353
3354static const struct proto_ops msg_ops = {
3355 .owner = THIS_MODULE,
3356 .family = AF_TIPC,
3357 .release = tipc_release,
3358 .bind = tipc_bind,
3359 .connect = tipc_connect,
3360 .socketpair = tipc_socketpair,
3361 .accept = sock_no_accept,
3362 .getname = tipc_getname,
3363 .poll = tipc_poll,
3364 .ioctl = tipc_ioctl,
3365 .listen = sock_no_listen,
3366 .shutdown = tipc_shutdown,
3367 .setsockopt = tipc_setsockopt,
3368 .getsockopt = tipc_getsockopt,
3369 .sendmsg = tipc_sendmsg,
3370 .recvmsg = tipc_recvmsg,
3371 .mmap = sock_no_mmap,
3372};
3373
3374static const struct proto_ops packet_ops = {
3375 .owner = THIS_MODULE,
3376 .family = AF_TIPC,
3377 .release = tipc_release,
3378 .bind = tipc_bind,
3379 .connect = tipc_connect,
3380 .socketpair = tipc_socketpair,
3381 .accept = tipc_accept,
3382 .getname = tipc_getname,
3383 .poll = tipc_poll,
3384 .ioctl = tipc_ioctl,
3385 .listen = tipc_listen,
3386 .shutdown = tipc_shutdown,
3387 .setsockopt = tipc_setsockopt,
3388 .getsockopt = tipc_getsockopt,
3389 .sendmsg = tipc_send_packet,
3390 .recvmsg = tipc_recvmsg,
3391 .mmap = sock_no_mmap,
3392};
3393
3394static const struct proto_ops stream_ops = {
3395 .owner = THIS_MODULE,
3396 .family = AF_TIPC,
3397 .release = tipc_release,
3398 .bind = tipc_bind,
3399 .connect = tipc_connect,
3400 .socketpair = tipc_socketpair,
3401 .accept = tipc_accept,
3402 .getname = tipc_getname,
3403 .poll = tipc_poll,
3404 .ioctl = tipc_ioctl,
3405 .listen = tipc_listen,
3406 .shutdown = tipc_shutdown,
3407 .setsockopt = tipc_setsockopt,
3408 .getsockopt = tipc_getsockopt,
3409 .sendmsg = tipc_sendstream,
3410 .recvmsg = tipc_recvstream,
3411 .mmap = sock_no_mmap,
3412};
3413
3414static const struct net_proto_family tipc_family_ops = {
3415 .owner = THIS_MODULE,
3416 .family = AF_TIPC,
3417 .create = tipc_sk_create
3418};
3419
3420static struct proto tipc_proto = {
3421 .name = "TIPC",
3422 .owner = THIS_MODULE,
3423 .obj_size = sizeof(struct tipc_sock),
3424 .sysctl_rmem = sysctl_tipc_rmem
3425};
3426
3427/**
3428 * tipc_socket_init - initialize TIPC socket interface
3429 *
3430 * Return: 0 on success, errno otherwise
3431 */
3432int tipc_socket_init(void)
3433{
3434 int res;
3435
3436 res = proto_register(&tipc_proto, 1);
3437 if (res) {
3438 pr_err("Failed to register TIPC protocol type\n");
3439 goto out;
3440 }
3441
3442 res = sock_register(&tipc_family_ops);
3443 if (res) {
3444 pr_err("Failed to register TIPC socket type\n");
3445 proto_unregister(&tipc_proto);
3446 goto out;
3447 }
3448 out:
3449 return res;
3450}
3451
3452/**
3453 * tipc_socket_stop - stop TIPC socket interface
3454 */
3455void tipc_socket_stop(void)
3456{
3457 sock_unregister(tipc_family_ops.family);
3458 proto_unregister(&tipc_proto);
3459}
3460
3461/* Caller should hold socket lock for the passed tipc socket. */
3462static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3463{
3464 u32 peer_node, peer_port;
3465 u32 conn_type, conn_instance;
3466 struct nlattr *nest;
3467
3468 peer_node = tsk_peer_node(tsk);
3469 peer_port = tsk_peer_port(tsk);
3470 conn_type = msg_nametype(&tsk->phdr);
3471 conn_instance = msg_nameinst(&tsk->phdr);
3472 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3473 if (!nest)
3474 return -EMSGSIZE;
3475
3476 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3477 goto msg_full;
3478 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3479 goto msg_full;
3480
3481 if (tsk->conn_addrtype != 0) {
3482 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3483 goto msg_full;
3484 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3485 goto msg_full;
3486 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3487 goto msg_full;
3488 }
3489 nla_nest_end(skb, nest);
3490
3491 return 0;
3492
3493msg_full:
3494 nla_nest_cancel(skb, nest);
3495
3496 return -EMSGSIZE;
3497}
3498
3499static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3500 *tsk)
3501{
3502 struct net *net = sock_net(skb->sk);
3503 struct sock *sk = &tsk->sk;
3504
3505 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3506 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3507 return -EMSGSIZE;
3508
3509 if (tipc_sk_connected(sk)) {
3510 if (__tipc_nl_add_sk_con(skb, tsk))
3511 return -EMSGSIZE;
3512 } else if (!list_empty(&tsk->publications)) {
3513 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3514 return -EMSGSIZE;
3515 }
3516 return 0;
3517}
3518
3519/* Caller should hold socket lock for the passed tipc socket. */
3520static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3521 struct tipc_sock *tsk)
3522{
3523 struct nlattr *attrs;
3524 void *hdr;
3525
3526 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3527 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3528 if (!hdr)
3529 goto msg_cancel;
3530
3531 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3532 if (!attrs)
3533 goto genlmsg_cancel;
3534
3535 if (__tipc_nl_add_sk_info(skb, tsk))
3536 goto attr_msg_cancel;
3537
3538 nla_nest_end(skb, attrs);
3539 genlmsg_end(skb, hdr);
3540
3541 return 0;
3542
3543attr_msg_cancel:
3544 nla_nest_cancel(skb, attrs);
3545genlmsg_cancel:
3546 genlmsg_cancel(skb, hdr);
3547msg_cancel:
3548 return -EMSGSIZE;
3549}
3550
3551int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3552 int (*skb_handler)(struct sk_buff *skb,
3553 struct netlink_callback *cb,
3554 struct tipc_sock *tsk))
3555{
3556 struct rhashtable_iter *iter = (void *)cb->args[4];
3557 struct tipc_sock *tsk;
3558 int err;
3559
3560 rhashtable_walk_start(iter);
3561 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3562 if (IS_ERR(tsk)) {
3563 if (PTR_ERR(tsk) == -EAGAIN)
3564 continue;
3565 break;
3566 }
3567
3568 sock_hold(&tsk->sk);
3569 rhashtable_walk_stop(iter);
3570 lock_sock(&tsk->sk);
3571 err = skb_handler(skb, cb, tsk);
3572 if (err) {
3573 release_sock(&tsk->sk);
3574 sock_put(&tsk->sk);
3575 goto out;
3576 }
3577 release_sock(&tsk->sk);
3578 rhashtable_walk_start(iter);
3579 sock_put(&tsk->sk);
3580 }
3581 rhashtable_walk_stop(iter);
3582out:
3583 return skb->len;
3584}
3585EXPORT_SYMBOL(tipc_nl_sk_walk);
3586
3587int tipc_dump_start(struct netlink_callback *cb)
3588{
3589 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3590}
3591EXPORT_SYMBOL(tipc_dump_start);
3592
3593int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3594{
3595 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3596 struct rhashtable_iter *iter = (void *)cb->args[4];
3597 struct tipc_net *tn = tipc_net(net);
3598
3599 if (!iter) {
3600 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3601 if (!iter)
3602 return -ENOMEM;
3603
3604 cb->args[4] = (long)iter;
3605 }
3606
3607 rhashtable_walk_enter(&tn->sk_rht, iter);
3608 return 0;
3609}
3610
3611int tipc_dump_done(struct netlink_callback *cb)
3612{
3613 struct rhashtable_iter *hti = (void *)cb->args[4];
3614
3615 rhashtable_walk_exit(hti);
3616 kfree(hti);
3617 return 0;
3618}
3619EXPORT_SYMBOL(tipc_dump_done);
3620
3621int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3622 struct tipc_sock *tsk, u32 sk_filter_state,
3623 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3624{
3625 struct sock *sk = &tsk->sk;
3626 struct nlattr *attrs;
3627 struct nlattr *stat;
3628
3629 /*filter response w.r.t sk_state*/
3630 if (!(sk_filter_state & (1 << sk->sk_state)))
3631 return 0;
3632
3633 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3634 if (!attrs)
3635 goto msg_cancel;
3636
3637 if (__tipc_nl_add_sk_info(skb, tsk))
3638 goto attr_msg_cancel;
3639
3640 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3641 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3642 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3643 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3644 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3645 sock_i_uid(sk))) ||
3646 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3647 tipc_diag_gen_cookie(sk),
3648 TIPC_NLA_SOCK_PAD))
3649 goto attr_msg_cancel;
3650
3651 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3652 if (!stat)
3653 goto attr_msg_cancel;
3654
3655 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3656 skb_queue_len(&sk->sk_receive_queue)) ||
3657 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3658 skb_queue_len(&sk->sk_write_queue)) ||
3659 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3660 atomic_read(&sk->sk_drops)))
3661 goto stat_msg_cancel;
3662
3663 if (tsk->cong_link_cnt &&
3664 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3665 goto stat_msg_cancel;
3666
3667 if (tsk_conn_cong(tsk) &&
3668 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3669 goto stat_msg_cancel;
3670
3671 nla_nest_end(skb, stat);
3672
3673 if (tsk->group)
3674 if (tipc_group_fill_sock_diag(tsk->group, skb))
3675 goto stat_msg_cancel;
3676
3677 nla_nest_end(skb, attrs);
3678
3679 return 0;
3680
3681stat_msg_cancel:
3682 nla_nest_cancel(skb, stat);
3683attr_msg_cancel:
3684 nla_nest_cancel(skb, attrs);
3685msg_cancel:
3686 return -EMSGSIZE;
3687}
3688EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3689
3690int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3691{
3692 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3693}
3694
3695/* Caller should hold socket lock for the passed tipc socket. */
3696static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3697 struct netlink_callback *cb,
3698 struct publication *publ)
3699{
3700 void *hdr;
3701 struct nlattr *attrs;
3702
3703 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3704 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3705 if (!hdr)
3706 goto msg_cancel;
3707
3708 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3709 if (!attrs)
3710 goto genlmsg_cancel;
3711
3712 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3713 goto attr_msg_cancel;
3714 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3715 goto attr_msg_cancel;
3716 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3717 goto attr_msg_cancel;
3718 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3719 goto attr_msg_cancel;
3720
3721 nla_nest_end(skb, attrs);
3722 genlmsg_end(skb, hdr);
3723
3724 return 0;
3725
3726attr_msg_cancel:
3727 nla_nest_cancel(skb, attrs);
3728genlmsg_cancel:
3729 genlmsg_cancel(skb, hdr);
3730msg_cancel:
3731 return -EMSGSIZE;
3732}
3733
3734/* Caller should hold socket lock for the passed tipc socket. */
3735static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3736 struct netlink_callback *cb,
3737 struct tipc_sock *tsk, u32 *last_publ)
3738{
3739 int err;
3740 struct publication *p;
3741
3742 if (*last_publ) {
3743 list_for_each_entry(p, &tsk->publications, binding_sock) {
3744 if (p->key == *last_publ)
3745 break;
3746 }
3747 if (list_entry_is_head(p, &tsk->publications, binding_sock)) {
3748 /* We never set seq or call nl_dump_check_consistent()
3749 * this means that setting prev_seq here will cause the
3750 * consistence check to fail in the netlink callback
3751 * handler. Resulting in the last NLMSG_DONE message
3752 * having the NLM_F_DUMP_INTR flag set.
3753 */
3754 cb->prev_seq = 1;
3755 *last_publ = 0;
3756 return -EPIPE;
3757 }
3758 } else {
3759 p = list_first_entry(&tsk->publications, struct publication,
3760 binding_sock);
3761 }
3762
3763 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3764 err = __tipc_nl_add_sk_publ(skb, cb, p);
3765 if (err) {
3766 *last_publ = p->key;
3767 return err;
3768 }
3769 }
3770 *last_publ = 0;
3771
3772 return 0;
3773}
3774
3775int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3776{
3777 int err;
3778 u32 tsk_portid = cb->args[0];
3779 u32 last_publ = cb->args[1];
3780 u32 done = cb->args[2];
3781 struct net *net = sock_net(skb->sk);
3782 struct tipc_sock *tsk;
3783
3784 if (!tsk_portid) {
3785 struct nlattr **attrs = genl_dumpit_info(cb)->info.attrs;
3786 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3787
3788 if (!attrs[TIPC_NLA_SOCK])
3789 return -EINVAL;
3790
3791 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3792 attrs[TIPC_NLA_SOCK],
3793 tipc_nl_sock_policy, NULL);
3794 if (err)
3795 return err;
3796
3797 if (!sock[TIPC_NLA_SOCK_REF])
3798 return -EINVAL;
3799
3800 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3801 }
3802
3803 if (done)
3804 return 0;
3805
3806 tsk = tipc_sk_lookup(net, tsk_portid);
3807 if (!tsk)
3808 return -EINVAL;
3809
3810 lock_sock(&tsk->sk);
3811 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3812 if (!err)
3813 done = 1;
3814 release_sock(&tsk->sk);
3815 sock_put(&tsk->sk);
3816
3817 cb->args[0] = tsk_portid;
3818 cb->args[1] = last_publ;
3819 cb->args[2] = done;
3820
3821 return skb->len;
3822}
3823
3824/**
3825 * tipc_sk_filtering - check if a socket should be traced
3826 * @sk: the socket to be examined
3827 *
3828 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3829 * (portid, sock type, name type, name lower, name upper)
3830 *
3831 * Return: true if the socket meets the socket tuple data
3832 * (value 0 = 'any') or when there is no tuple set (all = 0),
3833 * otherwise false
3834 */
3835bool tipc_sk_filtering(struct sock *sk)
3836{
3837 struct tipc_sock *tsk;
3838 struct publication *p;
3839 u32 _port, _sktype, _type, _lower, _upper;
3840 u32 type = 0, lower = 0, upper = 0;
3841
3842 if (!sk)
3843 return true;
3844
3845 tsk = tipc_sk(sk);
3846
3847 _port = sysctl_tipc_sk_filter[0];
3848 _sktype = sysctl_tipc_sk_filter[1];
3849 _type = sysctl_tipc_sk_filter[2];
3850 _lower = sysctl_tipc_sk_filter[3];
3851 _upper = sysctl_tipc_sk_filter[4];
3852
3853 if (!_port && !_sktype && !_type && !_lower && !_upper)
3854 return true;
3855
3856 if (_port)
3857 return (_port == tsk->portid);
3858
3859 if (_sktype && _sktype != sk->sk_type)
3860 return false;
3861
3862 if (tsk->published) {
3863 p = list_first_entry_or_null(&tsk->publications,
3864 struct publication, binding_sock);
3865 if (p) {
3866 type = p->sr.type;
3867 lower = p->sr.lower;
3868 upper = p->sr.upper;
3869 }
3870 }
3871
3872 if (!tipc_sk_type_connectionless(sk)) {
3873 type = msg_nametype(&tsk->phdr);
3874 lower = msg_nameinst(&tsk->phdr);
3875 upper = lower;
3876 }
3877
3878 if ((_type && _type != type) || (_lower && _lower != lower) ||
3879 (_upper && _upper != upper))
3880 return false;
3881
3882 return true;
3883}
3884
3885u32 tipc_sock_get_portid(struct sock *sk)
3886{
3887 return (sk) ? (tipc_sk(sk))->portid : 0;
3888}
3889
3890/**
3891 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3892 * both the rcv and backlog queues are considered
3893 * @sk: tipc sk to be checked
3894 * @skb: tipc msg to be checked
3895 *
3896 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3897 */
3898
3899bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3900{
3901 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3902 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3903 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3904
3905 return (qsize > lim * 90 / 100);
3906}
3907
3908/**
3909 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3910 * only the rcv queue is considered
3911 * @sk: tipc sk to be checked
3912 * @skb: tipc msg to be checked
3913 *
3914 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3915 */
3916
3917bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3918{
3919 unsigned int lim = rcvbuf_limit(sk, skb);
3920 unsigned int qsize = sk_rmem_alloc_get(sk);
3921
3922 return (qsize > lim * 90 / 100);
3923}
3924
3925/**
3926 * tipc_sk_dump - dump TIPC socket
3927 * @sk: tipc sk to be dumped
3928 * @dqueues: bitmask to decide if any socket queue to be dumped?
3929 * - TIPC_DUMP_NONE: don't dump socket queues
3930 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3931 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3932 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3933 * - TIPC_DUMP_ALL: dump all the socket queues above
3934 * @buf: returned buffer of dump data in format
3935 */
3936int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3937{
3938 int i = 0;
3939 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3940 u32 conn_type, conn_instance;
3941 struct tipc_sock *tsk;
3942 struct publication *p;
3943 bool tsk_connected;
3944
3945 if (!sk) {
3946 i += scnprintf(buf, sz, "sk data: (null)\n");
3947 return i;
3948 }
3949
3950 tsk = tipc_sk(sk);
3951 tsk_connected = !tipc_sk_type_connectionless(sk);
3952
3953 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3954 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3955 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3956 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3957 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3958 if (tsk_connected) {
3959 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3960 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3961 conn_type = msg_nametype(&tsk->phdr);
3962 conn_instance = msg_nameinst(&tsk->phdr);
3963 i += scnprintf(buf + i, sz - i, " %u", conn_type);
3964 i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3965 }
3966 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3967 if (tsk->published) {
3968 p = list_first_entry_or_null(&tsk->publications,
3969 struct publication, binding_sock);
3970 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3971 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3972 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3973 }
3974 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3975 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3976 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3977 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3978 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3979 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3980 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3981 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3982 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3983 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3984 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3985 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3986 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3987 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3988
3989 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3990 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3991 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3992 }
3993
3994 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3995 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3996 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
3997 }
3998
3999 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4000 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
4001 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4002 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4003 i += scnprintf(buf + i, sz - i, " tail ");
4004 i += tipc_skb_dump(sk->sk_backlog.tail, false,
4005 buf + i);
4006 }
4007 }
4008
4009 return i;
4010}
1/*
2 * net/tipc/socket.c: TIPC socket API
3 *
4 * Copyright (c) 2001-2007, 2012-2019, Ericsson AB
5 * Copyright (c) 2004-2008, 2010-2013, Wind River Systems
6 * Copyright (c) 2020-2021, Red Hat Inc
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions are met:
11 *
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the names of the copyright holders nor the names of its
18 * contributors may be used to endorse or promote products derived from
19 * this software without specific prior written permission.
20 *
21 * Alternatively, this software may be distributed under the terms of the
22 * GNU General Public License ("GPL") version 2 as published by the Free
23 * Software Foundation.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
36 */
37
38#include <linux/rhashtable.h>
39#include <linux/sched/signal.h>
40
41#include "core.h"
42#include "name_table.h"
43#include "node.h"
44#include "link.h"
45#include "name_distr.h"
46#include "socket.h"
47#include "bcast.h"
48#include "netlink.h"
49#include "group.h"
50#include "trace.h"
51
52#define NAGLE_START_INIT 4
53#define NAGLE_START_MAX 1024
54#define CONN_TIMEOUT_DEFAULT 8000 /* default connect timeout = 8s */
55#define CONN_PROBING_INTV msecs_to_jiffies(3600000) /* [ms] => 1 h */
56#define TIPC_MAX_PORT 0xffffffff
57#define TIPC_MIN_PORT 1
58#define TIPC_ACK_RATE 4 /* ACK at 1/4 of rcv window size */
59
60enum {
61 TIPC_LISTEN = TCP_LISTEN,
62 TIPC_ESTABLISHED = TCP_ESTABLISHED,
63 TIPC_OPEN = TCP_CLOSE,
64 TIPC_DISCONNECTING = TCP_CLOSE_WAIT,
65 TIPC_CONNECTING = TCP_SYN_SENT,
66};
67
68struct sockaddr_pair {
69 struct sockaddr_tipc sock;
70 struct sockaddr_tipc member;
71};
72
73/**
74 * struct tipc_sock - TIPC socket structure
75 * @sk: socket - interacts with 'port' and with user via the socket API
76 * @max_pkt: maximum packet size "hint" used when building messages sent by port
77 * @maxnagle: maximum size of msg which can be subject to nagle
78 * @portid: unique port identity in TIPC socket hash table
79 * @phdr: preformatted message header used when sending messages
80 * @cong_links: list of congested links
81 * @publications: list of publications for port
82 * @blocking_link: address of the congested link we are currently sleeping on
83 * @pub_count: total # of publications port has made during its lifetime
84 * @conn_timeout: the time we can wait for an unresponded setup request
85 * @probe_unacked: probe has not received ack yet
86 * @dupl_rcvcnt: number of bytes counted twice, in both backlog and rcv queue
87 * @cong_link_cnt: number of congested links
88 * @snt_unacked: # messages sent by socket, and not yet acked by peer
89 * @snd_win: send window size
90 * @peer_caps: peer capabilities mask
91 * @rcv_unacked: # messages read by user, but not yet acked back to peer
92 * @rcv_win: receive window size
93 * @peer: 'connected' peer for dgram/rdm
94 * @node: hash table node
95 * @mc_method: cookie for use between socket and broadcast layer
96 * @rcu: rcu struct for tipc_sock
97 * @group: TIPC communications group
98 * @oneway: message count in one direction (FIXME)
99 * @nagle_start: current nagle value
100 * @snd_backlog: send backlog count
101 * @msg_acc: messages accepted; used in managing backlog and nagle
102 * @pkt_cnt: TIPC socket packet count
103 * @expect_ack: whether this TIPC socket is expecting an ack
104 * @nodelay: setsockopt() TIPC_NODELAY setting
105 * @group_is_open: TIPC socket group is fully open (FIXME)
106 * @published: true if port has one or more associated names
107 * @conn_addrtype: address type used when establishing connection
108 */
109struct tipc_sock {
110 struct sock sk;
111 u32 max_pkt;
112 u32 maxnagle;
113 u32 portid;
114 struct tipc_msg phdr;
115 struct list_head cong_links;
116 struct list_head publications;
117 u32 pub_count;
118 atomic_t dupl_rcvcnt;
119 u16 conn_timeout;
120 bool probe_unacked;
121 u16 cong_link_cnt;
122 u16 snt_unacked;
123 u16 snd_win;
124 u16 peer_caps;
125 u16 rcv_unacked;
126 u16 rcv_win;
127 struct sockaddr_tipc peer;
128 struct rhash_head node;
129 struct tipc_mc_method mc_method;
130 struct rcu_head rcu;
131 struct tipc_group *group;
132 u32 oneway;
133 u32 nagle_start;
134 u16 snd_backlog;
135 u16 msg_acc;
136 u16 pkt_cnt;
137 bool expect_ack;
138 bool nodelay;
139 bool group_is_open;
140 bool published;
141 u8 conn_addrtype;
142};
143
144static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
145static void tipc_data_ready(struct sock *sk);
146static void tipc_write_space(struct sock *sk);
147static void tipc_sock_destruct(struct sock *sk);
148static int tipc_release(struct socket *sock);
149static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
150 bool kern);
151static void tipc_sk_timeout(struct timer_list *t);
152static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua);
153static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua);
154static int tipc_sk_leave(struct tipc_sock *tsk);
155static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid);
156static int tipc_sk_insert(struct tipc_sock *tsk);
157static void tipc_sk_remove(struct tipc_sock *tsk);
158static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz);
159static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dsz);
160static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack);
161static int tipc_wait_for_connect(struct socket *sock, long *timeo_p);
162
163static const struct proto_ops packet_ops;
164static const struct proto_ops stream_ops;
165static const struct proto_ops msg_ops;
166static struct proto tipc_proto;
167static const struct rhashtable_params tsk_rht_params;
168
169static u32 tsk_own_node(struct tipc_sock *tsk)
170{
171 return msg_prevnode(&tsk->phdr);
172}
173
174static u32 tsk_peer_node(struct tipc_sock *tsk)
175{
176 return msg_destnode(&tsk->phdr);
177}
178
179static u32 tsk_peer_port(struct tipc_sock *tsk)
180{
181 return msg_destport(&tsk->phdr);
182}
183
184static bool tsk_unreliable(struct tipc_sock *tsk)
185{
186 return msg_src_droppable(&tsk->phdr) != 0;
187}
188
189static void tsk_set_unreliable(struct tipc_sock *tsk, bool unreliable)
190{
191 msg_set_src_droppable(&tsk->phdr, unreliable ? 1 : 0);
192}
193
194static bool tsk_unreturnable(struct tipc_sock *tsk)
195{
196 return msg_dest_droppable(&tsk->phdr) != 0;
197}
198
199static void tsk_set_unreturnable(struct tipc_sock *tsk, bool unreturnable)
200{
201 msg_set_dest_droppable(&tsk->phdr, unreturnable ? 1 : 0);
202}
203
204static int tsk_importance(struct tipc_sock *tsk)
205{
206 return msg_importance(&tsk->phdr);
207}
208
209static struct tipc_sock *tipc_sk(const struct sock *sk)
210{
211 return container_of(sk, struct tipc_sock, sk);
212}
213
214int tsk_set_importance(struct sock *sk, int imp)
215{
216 if (imp > TIPC_CRITICAL_IMPORTANCE)
217 return -EINVAL;
218 msg_set_importance(&tipc_sk(sk)->phdr, (u32)imp);
219 return 0;
220}
221
222static bool tsk_conn_cong(struct tipc_sock *tsk)
223{
224 return tsk->snt_unacked > tsk->snd_win;
225}
226
227static u16 tsk_blocks(int len)
228{
229 return ((len / FLOWCTL_BLK_SZ) + 1);
230}
231
232/* tsk_blocks(): translate a buffer size in bytes to number of
233 * advertisable blocks, taking into account the ratio truesize(len)/len
234 * We can trust that this ratio is always < 4 for len >= FLOWCTL_BLK_SZ
235 */
236static u16 tsk_adv_blocks(int len)
237{
238 return len / FLOWCTL_BLK_SZ / 4;
239}
240
241/* tsk_inc(): increment counter for sent or received data
242 * - If block based flow control is not supported by peer we
243 * fall back to message based ditto, incrementing the counter
244 */
245static u16 tsk_inc(struct tipc_sock *tsk, int msglen)
246{
247 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
248 return ((msglen / FLOWCTL_BLK_SZ) + 1);
249 return 1;
250}
251
252/* tsk_set_nagle - enable/disable nagle property by manipulating maxnagle
253 */
254static void tsk_set_nagle(struct tipc_sock *tsk)
255{
256 struct sock *sk = &tsk->sk;
257
258 tsk->maxnagle = 0;
259 if (sk->sk_type != SOCK_STREAM)
260 return;
261 if (tsk->nodelay)
262 return;
263 if (!(tsk->peer_caps & TIPC_NAGLE))
264 return;
265 /* Limit node local buffer size to avoid receive queue overflow */
266 if (tsk->max_pkt == MAX_MSG_SIZE)
267 tsk->maxnagle = 1500;
268 else
269 tsk->maxnagle = tsk->max_pkt;
270}
271
272/**
273 * tsk_advance_rx_queue - discard first buffer in socket receive queue
274 * @sk: network socket
275 *
276 * Caller must hold socket lock
277 */
278static void tsk_advance_rx_queue(struct sock *sk)
279{
280 trace_tipc_sk_advance_rx(sk, NULL, TIPC_DUMP_SK_RCVQ, " ");
281 kfree_skb(__skb_dequeue(&sk->sk_receive_queue));
282}
283
284/* tipc_sk_respond() : send response message back to sender
285 */
286static void tipc_sk_respond(struct sock *sk, struct sk_buff *skb, int err)
287{
288 u32 selector;
289 u32 dnode;
290 u32 onode = tipc_own_addr(sock_net(sk));
291
292 if (!tipc_msg_reverse(onode, &skb, err))
293 return;
294
295 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE, "@sk_respond!");
296 dnode = msg_destnode(buf_msg(skb));
297 selector = msg_origport(buf_msg(skb));
298 tipc_node_xmit_skb(sock_net(sk), skb, dnode, selector);
299}
300
301/**
302 * tsk_rej_rx_queue - reject all buffers in socket receive queue
303 * @sk: network socket
304 * @error: response error code
305 *
306 * Caller must hold socket lock
307 */
308static void tsk_rej_rx_queue(struct sock *sk, int error)
309{
310 struct sk_buff *skb;
311
312 while ((skb = __skb_dequeue(&sk->sk_receive_queue)))
313 tipc_sk_respond(sk, skb, error);
314}
315
316static bool tipc_sk_connected(struct sock *sk)
317{
318 return sk->sk_state == TIPC_ESTABLISHED;
319}
320
321/* tipc_sk_type_connectionless - check if the socket is datagram socket
322 * @sk: socket
323 *
324 * Returns true if connection less, false otherwise
325 */
326static bool tipc_sk_type_connectionless(struct sock *sk)
327{
328 return sk->sk_type == SOCK_RDM || sk->sk_type == SOCK_DGRAM;
329}
330
331/* tsk_peer_msg - verify if message was sent by connected port's peer
332 *
333 * Handles cases where the node's network address has changed from
334 * the default of <0.0.0> to its configured setting.
335 */
336static bool tsk_peer_msg(struct tipc_sock *tsk, struct tipc_msg *msg)
337{
338 struct sock *sk = &tsk->sk;
339 u32 self = tipc_own_addr(sock_net(sk));
340 u32 peer_port = tsk_peer_port(tsk);
341 u32 orig_node, peer_node;
342
343 if (unlikely(!tipc_sk_connected(sk)))
344 return false;
345
346 if (unlikely(msg_origport(msg) != peer_port))
347 return false;
348
349 orig_node = msg_orignode(msg);
350 peer_node = tsk_peer_node(tsk);
351
352 if (likely(orig_node == peer_node))
353 return true;
354
355 if (!orig_node && peer_node == self)
356 return true;
357
358 if (!peer_node && orig_node == self)
359 return true;
360
361 return false;
362}
363
364/* tipc_set_sk_state - set the sk_state of the socket
365 * @sk: socket
366 *
367 * Caller must hold socket lock
368 *
369 * Returns 0 on success, errno otherwise
370 */
371static int tipc_set_sk_state(struct sock *sk, int state)
372{
373 int oldsk_state = sk->sk_state;
374 int res = -EINVAL;
375
376 switch (state) {
377 case TIPC_OPEN:
378 res = 0;
379 break;
380 case TIPC_LISTEN:
381 case TIPC_CONNECTING:
382 if (oldsk_state == TIPC_OPEN)
383 res = 0;
384 break;
385 case TIPC_ESTABLISHED:
386 if (oldsk_state == TIPC_CONNECTING ||
387 oldsk_state == TIPC_OPEN)
388 res = 0;
389 break;
390 case TIPC_DISCONNECTING:
391 if (oldsk_state == TIPC_CONNECTING ||
392 oldsk_state == TIPC_ESTABLISHED)
393 res = 0;
394 break;
395 }
396
397 if (!res)
398 sk->sk_state = state;
399
400 return res;
401}
402
403static int tipc_sk_sock_err(struct socket *sock, long *timeout)
404{
405 struct sock *sk = sock->sk;
406 int err = sock_error(sk);
407 int typ = sock->type;
408
409 if (err)
410 return err;
411 if (typ == SOCK_STREAM || typ == SOCK_SEQPACKET) {
412 if (sk->sk_state == TIPC_DISCONNECTING)
413 return -EPIPE;
414 else if (!tipc_sk_connected(sk))
415 return -ENOTCONN;
416 }
417 if (!*timeout)
418 return -EAGAIN;
419 if (signal_pending(current))
420 return sock_intr_errno(*timeout);
421
422 return 0;
423}
424
425#define tipc_wait_for_cond(sock_, timeo_, condition_) \
426({ \
427 DEFINE_WAIT_FUNC(wait_, woken_wake_function); \
428 struct sock *sk_; \
429 int rc_; \
430 \
431 while ((rc_ = !(condition_))) { \
432 /* coupled with smp_wmb() in tipc_sk_proto_rcv() */ \
433 smp_rmb(); \
434 sk_ = (sock_)->sk; \
435 rc_ = tipc_sk_sock_err((sock_), timeo_); \
436 if (rc_) \
437 break; \
438 add_wait_queue(sk_sleep(sk_), &wait_); \
439 release_sock(sk_); \
440 *(timeo_) = wait_woken(&wait_, TASK_INTERRUPTIBLE, *(timeo_)); \
441 sched_annotate_sleep(); \
442 lock_sock(sk_); \
443 remove_wait_queue(sk_sleep(sk_), &wait_); \
444 } \
445 rc_; \
446})
447
448/**
449 * tipc_sk_create - create a TIPC socket
450 * @net: network namespace (must be default network)
451 * @sock: pre-allocated socket structure
452 * @protocol: protocol indicator (must be 0)
453 * @kern: caused by kernel or by userspace?
454 *
455 * This routine creates additional data structures used by the TIPC socket,
456 * initializes them, and links them together.
457 *
458 * Return: 0 on success, errno otherwise
459 */
460static int tipc_sk_create(struct net *net, struct socket *sock,
461 int protocol, int kern)
462{
463 const struct proto_ops *ops;
464 struct sock *sk;
465 struct tipc_sock *tsk;
466 struct tipc_msg *msg;
467
468 /* Validate arguments */
469 if (unlikely(protocol != 0))
470 return -EPROTONOSUPPORT;
471
472 switch (sock->type) {
473 case SOCK_STREAM:
474 ops = &stream_ops;
475 break;
476 case SOCK_SEQPACKET:
477 ops = &packet_ops;
478 break;
479 case SOCK_DGRAM:
480 case SOCK_RDM:
481 ops = &msg_ops;
482 break;
483 default:
484 return -EPROTOTYPE;
485 }
486
487 /* Allocate socket's protocol area */
488 sk = sk_alloc(net, AF_TIPC, GFP_KERNEL, &tipc_proto, kern);
489 if (sk == NULL)
490 return -ENOMEM;
491
492 tsk = tipc_sk(sk);
493 tsk->max_pkt = MAX_PKT_DEFAULT;
494 tsk->maxnagle = 0;
495 tsk->nagle_start = NAGLE_START_INIT;
496 INIT_LIST_HEAD(&tsk->publications);
497 INIT_LIST_HEAD(&tsk->cong_links);
498 msg = &tsk->phdr;
499
500 /* Finish initializing socket data structures */
501 sock->ops = ops;
502 sock_init_data(sock, sk);
503 tipc_set_sk_state(sk, TIPC_OPEN);
504 if (tipc_sk_insert(tsk)) {
505 pr_warn("Socket create failed; port number exhausted\n");
506 return -EINVAL;
507 }
508
509 /* Ensure tsk is visible before we read own_addr. */
510 smp_mb();
511
512 tipc_msg_init(tipc_own_addr(net), msg, TIPC_LOW_IMPORTANCE,
513 TIPC_NAMED_MSG, NAMED_H_SIZE, 0);
514
515 msg_set_origport(msg, tsk->portid);
516 timer_setup(&sk->sk_timer, tipc_sk_timeout, 0);
517 sk->sk_shutdown = 0;
518 sk->sk_backlog_rcv = tipc_sk_backlog_rcv;
519 sk->sk_rcvbuf = sysctl_tipc_rmem[1];
520 sk->sk_data_ready = tipc_data_ready;
521 sk->sk_write_space = tipc_write_space;
522 sk->sk_destruct = tipc_sock_destruct;
523 tsk->conn_timeout = CONN_TIMEOUT_DEFAULT;
524 tsk->group_is_open = true;
525 atomic_set(&tsk->dupl_rcvcnt, 0);
526
527 /* Start out with safe limits until we receive an advertised window */
528 tsk->snd_win = tsk_adv_blocks(RCVBUF_MIN);
529 tsk->rcv_win = tsk->snd_win;
530
531 if (tipc_sk_type_connectionless(sk)) {
532 tsk_set_unreturnable(tsk, true);
533 if (sock->type == SOCK_DGRAM)
534 tsk_set_unreliable(tsk, true);
535 }
536 __skb_queue_head_init(&tsk->mc_method.deferredq);
537 trace_tipc_sk_create(sk, NULL, TIPC_DUMP_NONE, " ");
538 return 0;
539}
540
541static void tipc_sk_callback(struct rcu_head *head)
542{
543 struct tipc_sock *tsk = container_of(head, struct tipc_sock, rcu);
544
545 sock_put(&tsk->sk);
546}
547
548/* Caller should hold socket lock for the socket. */
549static void __tipc_shutdown(struct socket *sock, int error)
550{
551 struct sock *sk = sock->sk;
552 struct tipc_sock *tsk = tipc_sk(sk);
553 struct net *net = sock_net(sk);
554 long timeout = msecs_to_jiffies(CONN_TIMEOUT_DEFAULT);
555 u32 dnode = tsk_peer_node(tsk);
556 struct sk_buff *skb;
557
558 /* Avoid that hi-prio shutdown msgs bypass msgs in link wakeup queue */
559 tipc_wait_for_cond(sock, &timeout, (!tsk->cong_link_cnt &&
560 !tsk_conn_cong(tsk)));
561
562 /* Push out delayed messages if in Nagle mode */
563 tipc_sk_push_backlog(tsk, false);
564 /* Remove pending SYN */
565 __skb_queue_purge(&sk->sk_write_queue);
566
567 /* Remove partially received buffer if any */
568 skb = skb_peek(&sk->sk_receive_queue);
569 if (skb && TIPC_SKB_CB(skb)->bytes_read) {
570 __skb_unlink(skb, &sk->sk_receive_queue);
571 kfree_skb(skb);
572 }
573
574 /* Reject all unreceived messages if connectionless */
575 if (tipc_sk_type_connectionless(sk)) {
576 tsk_rej_rx_queue(sk, error);
577 return;
578 }
579
580 switch (sk->sk_state) {
581 case TIPC_CONNECTING:
582 case TIPC_ESTABLISHED:
583 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
584 tipc_node_remove_conn(net, dnode, tsk->portid);
585 /* Send a FIN+/- to its peer */
586 skb = __skb_dequeue(&sk->sk_receive_queue);
587 if (skb) {
588 __skb_queue_purge(&sk->sk_receive_queue);
589 tipc_sk_respond(sk, skb, error);
590 break;
591 }
592 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE,
593 TIPC_CONN_MSG, SHORT_H_SIZE, 0, dnode,
594 tsk_own_node(tsk), tsk_peer_port(tsk),
595 tsk->portid, error);
596 if (skb)
597 tipc_node_xmit_skb(net, skb, dnode, tsk->portid);
598 break;
599 case TIPC_LISTEN:
600 /* Reject all SYN messages */
601 tsk_rej_rx_queue(sk, error);
602 break;
603 default:
604 __skb_queue_purge(&sk->sk_receive_queue);
605 break;
606 }
607}
608
609/**
610 * tipc_release - destroy a TIPC socket
611 * @sock: socket to destroy
612 *
613 * This routine cleans up any messages that are still queued on the socket.
614 * For DGRAM and RDM socket types, all queued messages are rejected.
615 * For SEQPACKET and STREAM socket types, the first message is rejected
616 * and any others are discarded. (If the first message on a STREAM socket
617 * is partially-read, it is discarded and the next one is rejected instead.)
618 *
619 * NOTE: Rejected messages are not necessarily returned to the sender! They
620 * are returned or discarded according to the "destination droppable" setting
621 * specified for the message by the sender.
622 *
623 * Return: 0 on success, errno otherwise
624 */
625static int tipc_release(struct socket *sock)
626{
627 struct sock *sk = sock->sk;
628 struct tipc_sock *tsk;
629
630 /*
631 * Exit if socket isn't fully initialized (occurs when a failed accept()
632 * releases a pre-allocated child socket that was never used)
633 */
634 if (sk == NULL)
635 return 0;
636
637 tsk = tipc_sk(sk);
638 lock_sock(sk);
639
640 trace_tipc_sk_release(sk, NULL, TIPC_DUMP_ALL, " ");
641 __tipc_shutdown(sock, TIPC_ERR_NO_PORT);
642 sk->sk_shutdown = SHUTDOWN_MASK;
643 tipc_sk_leave(tsk);
644 tipc_sk_withdraw(tsk, NULL);
645 __skb_queue_purge(&tsk->mc_method.deferredq);
646 sk_stop_timer(sk, &sk->sk_timer);
647 tipc_sk_remove(tsk);
648
649 sock_orphan(sk);
650 /* Reject any messages that accumulated in backlog queue */
651 release_sock(sk);
652 tipc_dest_list_purge(&tsk->cong_links);
653 tsk->cong_link_cnt = 0;
654 call_rcu(&tsk->rcu, tipc_sk_callback);
655 sock->sk = NULL;
656
657 return 0;
658}
659
660/**
661 * __tipc_bind - associate or disassocate TIPC name(s) with a socket
662 * @sock: socket structure
663 * @skaddr: socket address describing name(s) and desired operation
664 * @alen: size of socket address data structure
665 *
666 * Name and name sequence binding are indicated using a positive scope value;
667 * a negative scope value unbinds the specified name. Specifying no name
668 * (i.e. a socket address length of 0) unbinds all names from the socket.
669 *
670 * Return: 0 on success, errno otherwise
671 *
672 * NOTE: This routine doesn't need to take the socket lock since it doesn't
673 * access any non-constant socket information.
674 */
675static int __tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
676{
677 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
678 struct tipc_sock *tsk = tipc_sk(sock->sk);
679 bool unbind = false;
680
681 if (unlikely(!alen))
682 return tipc_sk_withdraw(tsk, NULL);
683
684 if (ua->addrtype == TIPC_SERVICE_ADDR) {
685 ua->addrtype = TIPC_SERVICE_RANGE;
686 ua->sr.upper = ua->sr.lower;
687 }
688 if (ua->scope < 0) {
689 unbind = true;
690 ua->scope = -ua->scope;
691 }
692 /* Users may still use deprecated TIPC_ZONE_SCOPE */
693 if (ua->scope != TIPC_NODE_SCOPE)
694 ua->scope = TIPC_CLUSTER_SCOPE;
695
696 if (tsk->group)
697 return -EACCES;
698
699 if (unbind)
700 return tipc_sk_withdraw(tsk, ua);
701 return tipc_sk_publish(tsk, ua);
702}
703
704int tipc_sk_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
705{
706 int res;
707
708 lock_sock(sock->sk);
709 res = __tipc_bind(sock, skaddr, alen);
710 release_sock(sock->sk);
711 return res;
712}
713
714static int tipc_bind(struct socket *sock, struct sockaddr *skaddr, int alen)
715{
716 struct tipc_uaddr *ua = (struct tipc_uaddr *)skaddr;
717 u32 atype = ua->addrtype;
718
719 if (alen) {
720 if (!tipc_uaddr_valid(ua, alen))
721 return -EINVAL;
722 if (atype == TIPC_SOCKET_ADDR)
723 return -EAFNOSUPPORT;
724 if (ua->sr.type < TIPC_RESERVED_TYPES) {
725 pr_warn_once("Can't bind to reserved service type %u\n",
726 ua->sr.type);
727 return -EACCES;
728 }
729 }
730 return tipc_sk_bind(sock, skaddr, alen);
731}
732
733/**
734 * tipc_getname - get port ID of socket or peer socket
735 * @sock: socket structure
736 * @uaddr: area for returned socket address
737 * @peer: 0 = own ID, 1 = current peer ID, 2 = current/former peer ID
738 *
739 * Return: 0 on success, errno otherwise
740 *
741 * NOTE: This routine doesn't need to take the socket lock since it only
742 * accesses socket information that is unchanging (or which changes in
743 * a completely predictable manner).
744 */
745static int tipc_getname(struct socket *sock, struct sockaddr *uaddr,
746 int peer)
747{
748 struct sockaddr_tipc *addr = (struct sockaddr_tipc *)uaddr;
749 struct sock *sk = sock->sk;
750 struct tipc_sock *tsk = tipc_sk(sk);
751
752 memset(addr, 0, sizeof(*addr));
753 if (peer) {
754 if ((!tipc_sk_connected(sk)) &&
755 ((peer != 2) || (sk->sk_state != TIPC_DISCONNECTING)))
756 return -ENOTCONN;
757 addr->addr.id.ref = tsk_peer_port(tsk);
758 addr->addr.id.node = tsk_peer_node(tsk);
759 } else {
760 addr->addr.id.ref = tsk->portid;
761 addr->addr.id.node = tipc_own_addr(sock_net(sk));
762 }
763
764 addr->addrtype = TIPC_SOCKET_ADDR;
765 addr->family = AF_TIPC;
766 addr->scope = 0;
767 addr->addr.name.domain = 0;
768
769 return sizeof(*addr);
770}
771
772/**
773 * tipc_poll - read and possibly block on pollmask
774 * @file: file structure associated with the socket
775 * @sock: socket for which to calculate the poll bits
776 * @wait: ???
777 *
778 * Return: pollmask value
779 *
780 * COMMENTARY:
781 * It appears that the usual socket locking mechanisms are not useful here
782 * since the pollmask info is potentially out-of-date the moment this routine
783 * exits. TCP and other protocols seem to rely on higher level poll routines
784 * to handle any preventable race conditions, so TIPC will do the same ...
785 *
786 * IMPORTANT: The fact that a read or write operation is indicated does NOT
787 * imply that the operation will succeed, merely that it should be performed
788 * and will not block.
789 */
790static __poll_t tipc_poll(struct file *file, struct socket *sock,
791 poll_table *wait)
792{
793 struct sock *sk = sock->sk;
794 struct tipc_sock *tsk = tipc_sk(sk);
795 __poll_t revents = 0;
796
797 sock_poll_wait(file, sock, wait);
798 trace_tipc_sk_poll(sk, NULL, TIPC_DUMP_ALL, " ");
799
800 if (sk->sk_shutdown & RCV_SHUTDOWN)
801 revents |= EPOLLRDHUP | EPOLLIN | EPOLLRDNORM;
802 if (sk->sk_shutdown == SHUTDOWN_MASK)
803 revents |= EPOLLHUP;
804
805 switch (sk->sk_state) {
806 case TIPC_ESTABLISHED:
807 if (!tsk->cong_link_cnt && !tsk_conn_cong(tsk))
808 revents |= EPOLLOUT;
809 fallthrough;
810 case TIPC_LISTEN:
811 case TIPC_CONNECTING:
812 if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
813 revents |= EPOLLIN | EPOLLRDNORM;
814 break;
815 case TIPC_OPEN:
816 if (tsk->group_is_open && !tsk->cong_link_cnt)
817 revents |= EPOLLOUT;
818 if (!tipc_sk_type_connectionless(sk))
819 break;
820 if (skb_queue_empty_lockless(&sk->sk_receive_queue))
821 break;
822 revents |= EPOLLIN | EPOLLRDNORM;
823 break;
824 case TIPC_DISCONNECTING:
825 revents = EPOLLIN | EPOLLRDNORM | EPOLLHUP;
826 break;
827 }
828 return revents;
829}
830
831/**
832 * tipc_sendmcast - send multicast message
833 * @sock: socket structure
834 * @ua: destination address struct
835 * @msg: message to send
836 * @dlen: length of data to send
837 * @timeout: timeout to wait for wakeup
838 *
839 * Called from function tipc_sendmsg(), which has done all sanity checks
840 * Return: the number of bytes sent on success, or errno
841 */
842static int tipc_sendmcast(struct socket *sock, struct tipc_uaddr *ua,
843 struct msghdr *msg, size_t dlen, long timeout)
844{
845 struct sock *sk = sock->sk;
846 struct tipc_sock *tsk = tipc_sk(sk);
847 struct tipc_msg *hdr = &tsk->phdr;
848 struct net *net = sock_net(sk);
849 int mtu = tipc_bcast_get_mtu(net);
850 struct sk_buff_head pkts;
851 struct tipc_nlist dsts;
852 int rc;
853
854 if (tsk->group)
855 return -EACCES;
856
857 /* Block or return if any destination link is congested */
858 rc = tipc_wait_for_cond(sock, &timeout, !tsk->cong_link_cnt);
859 if (unlikely(rc))
860 return rc;
861
862 /* Lookup destination nodes */
863 tipc_nlist_init(&dsts, tipc_own_addr(net));
864 tipc_nametbl_lookup_mcast_nodes(net, ua, &dsts);
865 if (!dsts.local && !dsts.remote)
866 return -EHOSTUNREACH;
867
868 /* Build message header */
869 msg_set_type(hdr, TIPC_MCAST_MSG);
870 msg_set_hdr_sz(hdr, MCAST_H_SIZE);
871 msg_set_lookup_scope(hdr, TIPC_CLUSTER_SCOPE);
872 msg_set_destport(hdr, 0);
873 msg_set_destnode(hdr, 0);
874 msg_set_nametype(hdr, ua->sr.type);
875 msg_set_namelower(hdr, ua->sr.lower);
876 msg_set_nameupper(hdr, ua->sr.upper);
877
878 /* Build message as chain of buffers */
879 __skb_queue_head_init(&pkts);
880 rc = tipc_msg_build(hdr, msg, 0, dlen, mtu, &pkts);
881
882 /* Send message if build was successful */
883 if (unlikely(rc == dlen)) {
884 trace_tipc_sk_sendmcast(sk, skb_peek(&pkts),
885 TIPC_DUMP_SK_SNDQ, " ");
886 rc = tipc_mcast_xmit(net, &pkts, &tsk->mc_method, &dsts,
887 &tsk->cong_link_cnt);
888 }
889
890 tipc_nlist_purge(&dsts);
891
892 return rc ? rc : dlen;
893}
894
895/**
896 * tipc_send_group_msg - send a message to a member in the group
897 * @net: network namespace
898 * @tsk: tipc socket
899 * @m: message to send
900 * @mb: group member
901 * @dnode: destination node
902 * @dport: destination port
903 * @dlen: total length of message data
904 */
905static int tipc_send_group_msg(struct net *net, struct tipc_sock *tsk,
906 struct msghdr *m, struct tipc_member *mb,
907 u32 dnode, u32 dport, int dlen)
908{
909 u16 bc_snd_nxt = tipc_group_bc_snd_nxt(tsk->group);
910 struct tipc_mc_method *method = &tsk->mc_method;
911 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
912 struct tipc_msg *hdr = &tsk->phdr;
913 struct sk_buff_head pkts;
914 int mtu, rc;
915
916 /* Complete message header */
917 msg_set_type(hdr, TIPC_GRP_UCAST_MSG);
918 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
919 msg_set_destport(hdr, dport);
920 msg_set_destnode(hdr, dnode);
921 msg_set_grp_bc_seqno(hdr, bc_snd_nxt);
922
923 /* Build message as chain of buffers */
924 __skb_queue_head_init(&pkts);
925 mtu = tipc_node_get_mtu(net, dnode, tsk->portid, false);
926 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
927 if (unlikely(rc != dlen))
928 return rc;
929
930 /* Send message */
931 rc = tipc_node_xmit(net, &pkts, dnode, tsk->portid);
932 if (unlikely(rc == -ELINKCONG)) {
933 tipc_dest_push(&tsk->cong_links, dnode, 0);
934 tsk->cong_link_cnt++;
935 }
936
937 /* Update send window */
938 tipc_group_update_member(mb, blks);
939
940 /* A broadcast sent within next EXPIRE period must follow same path */
941 method->rcast = true;
942 method->mandatory = true;
943 return dlen;
944}
945
946/**
947 * tipc_send_group_unicast - send message to a member in the group
948 * @sock: socket structure
949 * @m: message to send
950 * @dlen: total length of message data
951 * @timeout: timeout to wait for wakeup
952 *
953 * Called from function tipc_sendmsg(), which has done all sanity checks
954 * Return: the number of bytes sent on success, or errno
955 */
956static int tipc_send_group_unicast(struct socket *sock, struct msghdr *m,
957 int dlen, long timeout)
958{
959 struct sock *sk = sock->sk;
960 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
961 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
962 struct tipc_sock *tsk = tipc_sk(sk);
963 struct net *net = sock_net(sk);
964 struct tipc_member *mb = NULL;
965 u32 node, port;
966 int rc;
967
968 node = ua->sk.node;
969 port = ua->sk.ref;
970 if (!port && !node)
971 return -EHOSTUNREACH;
972
973 /* Block or return if destination link or member is congested */
974 rc = tipc_wait_for_cond(sock, &timeout,
975 !tipc_dest_find(&tsk->cong_links, node, 0) &&
976 tsk->group &&
977 !tipc_group_cong(tsk->group, node, port, blks,
978 &mb));
979 if (unlikely(rc))
980 return rc;
981
982 if (unlikely(!mb))
983 return -EHOSTUNREACH;
984
985 rc = tipc_send_group_msg(net, tsk, m, mb, node, port, dlen);
986
987 return rc ? rc : dlen;
988}
989
990/**
991 * tipc_send_group_anycast - send message to any member with given identity
992 * @sock: socket structure
993 * @m: message to send
994 * @dlen: total length of message data
995 * @timeout: timeout to wait for wakeup
996 *
997 * Called from function tipc_sendmsg(), which has done all sanity checks
998 * Return: the number of bytes sent on success, or errno
999 */
1000static int tipc_send_group_anycast(struct socket *sock, struct msghdr *m,
1001 int dlen, long timeout)
1002{
1003 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1004 struct sock *sk = sock->sk;
1005 struct tipc_sock *tsk = tipc_sk(sk);
1006 struct list_head *cong_links = &tsk->cong_links;
1007 int blks = tsk_blocks(GROUP_H_SIZE + dlen);
1008 struct tipc_msg *hdr = &tsk->phdr;
1009 struct tipc_member *first = NULL;
1010 struct tipc_member *mbr = NULL;
1011 struct net *net = sock_net(sk);
1012 u32 node, port, exclude;
1013 struct list_head dsts;
1014 int lookups = 0;
1015 int dstcnt, rc;
1016 bool cong;
1017
1018 INIT_LIST_HEAD(&dsts);
1019 ua->sa.type = msg_nametype(hdr);
1020 ua->scope = msg_lookup_scope(hdr);
1021
1022 while (++lookups < 4) {
1023 exclude = tipc_group_exclude(tsk->group);
1024
1025 first = NULL;
1026
1027 /* Look for a non-congested destination member, if any */
1028 while (1) {
1029 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt,
1030 exclude, false))
1031 return -EHOSTUNREACH;
1032 tipc_dest_pop(&dsts, &node, &port);
1033 cong = tipc_group_cong(tsk->group, node, port, blks,
1034 &mbr);
1035 if (!cong)
1036 break;
1037 if (mbr == first)
1038 break;
1039 if (!first)
1040 first = mbr;
1041 }
1042
1043 /* Start over if destination was not in member list */
1044 if (unlikely(!mbr))
1045 continue;
1046
1047 if (likely(!cong && !tipc_dest_find(cong_links, node, 0)))
1048 break;
1049
1050 /* Block or return if destination link or member is congested */
1051 rc = tipc_wait_for_cond(sock, &timeout,
1052 !tipc_dest_find(cong_links, node, 0) &&
1053 tsk->group &&
1054 !tipc_group_cong(tsk->group, node, port,
1055 blks, &mbr));
1056 if (unlikely(rc))
1057 return rc;
1058
1059 /* Send, unless destination disappeared while waiting */
1060 if (likely(mbr))
1061 break;
1062 }
1063
1064 if (unlikely(lookups >= 4))
1065 return -EHOSTUNREACH;
1066
1067 rc = tipc_send_group_msg(net, tsk, m, mbr, node, port, dlen);
1068
1069 return rc ? rc : dlen;
1070}
1071
1072/**
1073 * tipc_send_group_bcast - send message to all members in communication group
1074 * @sock: socket structure
1075 * @m: message to send
1076 * @dlen: total length of message data
1077 * @timeout: timeout to wait for wakeup
1078 *
1079 * Called from function tipc_sendmsg(), which has done all sanity checks
1080 * Return: the number of bytes sent on success, or errno
1081 */
1082static int tipc_send_group_bcast(struct socket *sock, struct msghdr *m,
1083 int dlen, long timeout)
1084{
1085 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1086 struct sock *sk = sock->sk;
1087 struct net *net = sock_net(sk);
1088 struct tipc_sock *tsk = tipc_sk(sk);
1089 struct tipc_nlist *dsts;
1090 struct tipc_mc_method *method = &tsk->mc_method;
1091 bool ack = method->mandatory && method->rcast;
1092 int blks = tsk_blocks(MCAST_H_SIZE + dlen);
1093 struct tipc_msg *hdr = &tsk->phdr;
1094 int mtu = tipc_bcast_get_mtu(net);
1095 struct sk_buff_head pkts;
1096 int rc = -EHOSTUNREACH;
1097
1098 /* Block or return if any destination link or member is congested */
1099 rc = tipc_wait_for_cond(sock, &timeout,
1100 !tsk->cong_link_cnt && tsk->group &&
1101 !tipc_group_bc_cong(tsk->group, blks));
1102 if (unlikely(rc))
1103 return rc;
1104
1105 dsts = tipc_group_dests(tsk->group);
1106 if (!dsts->local && !dsts->remote)
1107 return -EHOSTUNREACH;
1108
1109 /* Complete message header */
1110 if (ua) {
1111 msg_set_type(hdr, TIPC_GRP_MCAST_MSG);
1112 msg_set_nameinst(hdr, ua->sa.instance);
1113 } else {
1114 msg_set_type(hdr, TIPC_GRP_BCAST_MSG);
1115 msg_set_nameinst(hdr, 0);
1116 }
1117 msg_set_hdr_sz(hdr, GROUP_H_SIZE);
1118 msg_set_destport(hdr, 0);
1119 msg_set_destnode(hdr, 0);
1120 msg_set_grp_bc_seqno(hdr, tipc_group_bc_snd_nxt(tsk->group));
1121
1122 /* Avoid getting stuck with repeated forced replicasts */
1123 msg_set_grp_bc_ack_req(hdr, ack);
1124
1125 /* Build message as chain of buffers */
1126 __skb_queue_head_init(&pkts);
1127 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1128 if (unlikely(rc != dlen))
1129 return rc;
1130
1131 /* Send message */
1132 rc = tipc_mcast_xmit(net, &pkts, method, dsts, &tsk->cong_link_cnt);
1133 if (unlikely(rc))
1134 return rc;
1135
1136 /* Update broadcast sequence number and send windows */
1137 tipc_group_update_bc_members(tsk->group, blks, ack);
1138
1139 /* Broadcast link is now free to choose method for next broadcast */
1140 method->mandatory = false;
1141 method->expires = jiffies;
1142
1143 return dlen;
1144}
1145
1146/**
1147 * tipc_send_group_mcast - send message to all members with given identity
1148 * @sock: socket structure
1149 * @m: message to send
1150 * @dlen: total length of message data
1151 * @timeout: timeout to wait for wakeup
1152 *
1153 * Called from function tipc_sendmsg(), which has done all sanity checks
1154 * Return: the number of bytes sent on success, or errno
1155 */
1156static int tipc_send_group_mcast(struct socket *sock, struct msghdr *m,
1157 int dlen, long timeout)
1158{
1159 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1160 struct sock *sk = sock->sk;
1161 struct tipc_sock *tsk = tipc_sk(sk);
1162 struct tipc_group *grp = tsk->group;
1163 struct tipc_msg *hdr = &tsk->phdr;
1164 struct net *net = sock_net(sk);
1165 struct list_head dsts;
1166 u32 dstcnt, exclude;
1167
1168 INIT_LIST_HEAD(&dsts);
1169 ua->sa.type = msg_nametype(hdr);
1170 ua->scope = msg_lookup_scope(hdr);
1171 exclude = tipc_group_exclude(grp);
1172
1173 if (!tipc_nametbl_lookup_group(net, ua, &dsts, &dstcnt, exclude, true))
1174 return -EHOSTUNREACH;
1175
1176 if (dstcnt == 1) {
1177 tipc_dest_pop(&dsts, &ua->sk.node, &ua->sk.ref);
1178 return tipc_send_group_unicast(sock, m, dlen, timeout);
1179 }
1180
1181 tipc_dest_list_purge(&dsts);
1182 return tipc_send_group_bcast(sock, m, dlen, timeout);
1183}
1184
1185/**
1186 * tipc_sk_mcast_rcv - Deliver multicast messages to all destination sockets
1187 * @net: the associated network namespace
1188 * @arrvq: queue with arriving messages, to be cloned after destination lookup
1189 * @inputq: queue with cloned messages, delivered to socket after dest lookup
1190 *
1191 * Multi-threaded: parallel calls with reference to same queues may occur
1192 */
1193void tipc_sk_mcast_rcv(struct net *net, struct sk_buff_head *arrvq,
1194 struct sk_buff_head *inputq)
1195{
1196 u32 self = tipc_own_addr(net);
1197 struct sk_buff *skb, *_skb;
1198 u32 portid, onode;
1199 struct sk_buff_head tmpq;
1200 struct list_head dports;
1201 struct tipc_msg *hdr;
1202 struct tipc_uaddr ua;
1203 int user, mtyp, hlen;
1204
1205 __skb_queue_head_init(&tmpq);
1206 INIT_LIST_HEAD(&dports);
1207 ua.addrtype = TIPC_SERVICE_RANGE;
1208
1209 /* tipc_skb_peek() increments the head skb's reference counter */
1210 skb = tipc_skb_peek(arrvq, &inputq->lock);
1211 for (; skb; skb = tipc_skb_peek(arrvq, &inputq->lock)) {
1212 hdr = buf_msg(skb);
1213 user = msg_user(hdr);
1214 mtyp = msg_type(hdr);
1215 hlen = skb_headroom(skb) + msg_hdr_sz(hdr);
1216 onode = msg_orignode(hdr);
1217 ua.sr.type = msg_nametype(hdr);
1218 ua.sr.lower = msg_namelower(hdr);
1219 ua.sr.upper = msg_nameupper(hdr);
1220 if (onode == self)
1221 ua.scope = TIPC_ANY_SCOPE;
1222 else
1223 ua.scope = TIPC_CLUSTER_SCOPE;
1224
1225 if (mtyp == TIPC_GRP_UCAST_MSG || user == GROUP_PROTOCOL) {
1226 spin_lock_bh(&inputq->lock);
1227 if (skb_peek(arrvq) == skb) {
1228 __skb_dequeue(arrvq);
1229 __skb_queue_tail(inputq, skb);
1230 }
1231 kfree_skb(skb);
1232 spin_unlock_bh(&inputq->lock);
1233 continue;
1234 }
1235
1236 /* Group messages require exact scope match */
1237 if (msg_in_group(hdr)) {
1238 ua.sr.lower = 0;
1239 ua.sr.upper = ~0;
1240 ua.scope = msg_lookup_scope(hdr);
1241 }
1242
1243 /* Create destination port list: */
1244 tipc_nametbl_lookup_mcast_sockets(net, &ua, &dports);
1245
1246 /* Clone message per destination */
1247 while (tipc_dest_pop(&dports, NULL, &portid)) {
1248 _skb = __pskb_copy(skb, hlen, GFP_ATOMIC);
1249 if (_skb) {
1250 msg_set_destport(buf_msg(_skb), portid);
1251 __skb_queue_tail(&tmpq, _skb);
1252 continue;
1253 }
1254 pr_warn("Failed to clone mcast rcv buffer\n");
1255 }
1256 /* Append clones to inputq only if skb is still head of arrvq */
1257 spin_lock_bh(&inputq->lock);
1258 if (skb_peek(arrvq) == skb) {
1259 skb_queue_splice_tail_init(&tmpq, inputq);
1260 /* Decrement the skb's refcnt */
1261 kfree_skb(__skb_dequeue(arrvq));
1262 }
1263 spin_unlock_bh(&inputq->lock);
1264 __skb_queue_purge(&tmpq);
1265 kfree_skb(skb);
1266 }
1267 tipc_sk_rcv(net, inputq);
1268}
1269
1270/* tipc_sk_push_backlog(): send accumulated buffers in socket write queue
1271 * when socket is in Nagle mode
1272 */
1273static void tipc_sk_push_backlog(struct tipc_sock *tsk, bool nagle_ack)
1274{
1275 struct sk_buff_head *txq = &tsk->sk.sk_write_queue;
1276 struct sk_buff *skb = skb_peek_tail(txq);
1277 struct net *net = sock_net(&tsk->sk);
1278 u32 dnode = tsk_peer_node(tsk);
1279 int rc;
1280
1281 if (nagle_ack) {
1282 tsk->pkt_cnt += skb_queue_len(txq);
1283 if (!tsk->pkt_cnt || tsk->msg_acc / tsk->pkt_cnt < 2) {
1284 tsk->oneway = 0;
1285 if (tsk->nagle_start < NAGLE_START_MAX)
1286 tsk->nagle_start *= 2;
1287 tsk->expect_ack = false;
1288 pr_debug("tsk %10u: bad nagle %u -> %u, next start %u!\n",
1289 tsk->portid, tsk->msg_acc, tsk->pkt_cnt,
1290 tsk->nagle_start);
1291 } else {
1292 tsk->nagle_start = NAGLE_START_INIT;
1293 if (skb) {
1294 msg_set_ack_required(buf_msg(skb));
1295 tsk->expect_ack = true;
1296 } else {
1297 tsk->expect_ack = false;
1298 }
1299 }
1300 tsk->msg_acc = 0;
1301 tsk->pkt_cnt = 0;
1302 }
1303
1304 if (!skb || tsk->cong_link_cnt)
1305 return;
1306
1307 /* Do not send SYN again after congestion */
1308 if (msg_is_syn(buf_msg(skb)))
1309 return;
1310
1311 if (tsk->msg_acc)
1312 tsk->pkt_cnt += skb_queue_len(txq);
1313 tsk->snt_unacked += tsk->snd_backlog;
1314 tsk->snd_backlog = 0;
1315 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1316 if (rc == -ELINKCONG)
1317 tsk->cong_link_cnt = 1;
1318}
1319
1320/**
1321 * tipc_sk_conn_proto_rcv - receive a connection mng protocol message
1322 * @tsk: receiving socket
1323 * @skb: pointer to message buffer.
1324 * @inputq: buffer list containing the buffers
1325 * @xmitq: output message area
1326 */
1327static void tipc_sk_conn_proto_rcv(struct tipc_sock *tsk, struct sk_buff *skb,
1328 struct sk_buff_head *inputq,
1329 struct sk_buff_head *xmitq)
1330{
1331 struct tipc_msg *hdr = buf_msg(skb);
1332 u32 onode = tsk_own_node(tsk);
1333 struct sock *sk = &tsk->sk;
1334 int mtyp = msg_type(hdr);
1335 bool was_cong;
1336
1337 /* Ignore if connection cannot be validated: */
1338 if (!tsk_peer_msg(tsk, hdr)) {
1339 trace_tipc_sk_drop_msg(sk, skb, TIPC_DUMP_NONE, "@proto_rcv!");
1340 goto exit;
1341 }
1342
1343 if (unlikely(msg_errcode(hdr))) {
1344 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
1345 tipc_node_remove_conn(sock_net(sk), tsk_peer_node(tsk),
1346 tsk_peer_port(tsk));
1347 sk->sk_state_change(sk);
1348
1349 /* State change is ignored if socket already awake,
1350 * - convert msg to abort msg and add to inqueue
1351 */
1352 msg_set_user(hdr, TIPC_CRITICAL_IMPORTANCE);
1353 msg_set_type(hdr, TIPC_CONN_MSG);
1354 msg_set_size(hdr, BASIC_H_SIZE);
1355 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1356 __skb_queue_tail(inputq, skb);
1357 return;
1358 }
1359
1360 tsk->probe_unacked = false;
1361
1362 if (mtyp == CONN_PROBE) {
1363 msg_set_type(hdr, CONN_PROBE_REPLY);
1364 if (tipc_msg_reverse(onode, &skb, TIPC_OK))
1365 __skb_queue_tail(xmitq, skb);
1366 return;
1367 } else if (mtyp == CONN_ACK) {
1368 was_cong = tsk_conn_cong(tsk);
1369 tipc_sk_push_backlog(tsk, msg_nagle_ack(hdr));
1370 tsk->snt_unacked -= msg_conn_ack(hdr);
1371 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1372 tsk->snd_win = msg_adv_win(hdr);
1373 if (was_cong && !tsk_conn_cong(tsk))
1374 sk->sk_write_space(sk);
1375 } else if (mtyp != CONN_PROBE_REPLY) {
1376 pr_warn("Received unknown CONN_PROTO msg\n");
1377 }
1378exit:
1379 kfree_skb(skb);
1380}
1381
1382/**
1383 * tipc_sendmsg - send message in connectionless manner
1384 * @sock: socket structure
1385 * @m: message to send
1386 * @dsz: amount of user data to be sent
1387 *
1388 * Message must have an destination specified explicitly.
1389 * Used for SOCK_RDM and SOCK_DGRAM messages,
1390 * and for 'SYN' messages on SOCK_SEQPACKET and SOCK_STREAM connections.
1391 * (Note: 'SYN+' is prohibited on SOCK_STREAM.)
1392 *
1393 * Return: the number of bytes sent on success, or errno otherwise
1394 */
1395static int tipc_sendmsg(struct socket *sock,
1396 struct msghdr *m, size_t dsz)
1397{
1398 struct sock *sk = sock->sk;
1399 int ret;
1400
1401 lock_sock(sk);
1402 ret = __tipc_sendmsg(sock, m, dsz);
1403 release_sock(sk);
1404
1405 return ret;
1406}
1407
1408static int __tipc_sendmsg(struct socket *sock, struct msghdr *m, size_t dlen)
1409{
1410 struct sock *sk = sock->sk;
1411 struct net *net = sock_net(sk);
1412 struct tipc_sock *tsk = tipc_sk(sk);
1413 struct tipc_uaddr *ua = (struct tipc_uaddr *)m->msg_name;
1414 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1415 struct list_head *clinks = &tsk->cong_links;
1416 bool syn = !tipc_sk_type_connectionless(sk);
1417 struct tipc_group *grp = tsk->group;
1418 struct tipc_msg *hdr = &tsk->phdr;
1419 struct tipc_socket_addr skaddr;
1420 struct sk_buff_head pkts;
1421 int atype, mtu, rc;
1422
1423 if (unlikely(dlen > TIPC_MAX_USER_MSG_SIZE))
1424 return -EMSGSIZE;
1425
1426 if (ua) {
1427 if (!tipc_uaddr_valid(ua, m->msg_namelen))
1428 return -EINVAL;
1429 atype = ua->addrtype;
1430 }
1431
1432 /* If socket belongs to a communication group follow other paths */
1433 if (grp) {
1434 if (!ua)
1435 return tipc_send_group_bcast(sock, m, dlen, timeout);
1436 if (atype == TIPC_SERVICE_ADDR)
1437 return tipc_send_group_anycast(sock, m, dlen, timeout);
1438 if (atype == TIPC_SOCKET_ADDR)
1439 return tipc_send_group_unicast(sock, m, dlen, timeout);
1440 if (atype == TIPC_SERVICE_RANGE)
1441 return tipc_send_group_mcast(sock, m, dlen, timeout);
1442 return -EINVAL;
1443 }
1444
1445 if (!ua) {
1446 ua = (struct tipc_uaddr *)&tsk->peer;
1447 if (!syn && ua->family != AF_TIPC)
1448 return -EDESTADDRREQ;
1449 atype = ua->addrtype;
1450 }
1451
1452 if (unlikely(syn)) {
1453 if (sk->sk_state == TIPC_LISTEN)
1454 return -EPIPE;
1455 if (sk->sk_state != TIPC_OPEN)
1456 return -EISCONN;
1457 if (tsk->published)
1458 return -EOPNOTSUPP;
1459 if (atype == TIPC_SERVICE_ADDR)
1460 tsk->conn_addrtype = atype;
1461 msg_set_syn(hdr, 1);
1462 }
1463
1464 /* Determine destination */
1465 if (atype == TIPC_SERVICE_RANGE) {
1466 return tipc_sendmcast(sock, ua, m, dlen, timeout);
1467 } else if (atype == TIPC_SERVICE_ADDR) {
1468 skaddr.node = ua->lookup_node;
1469 ua->scope = tipc_node2scope(skaddr.node);
1470 if (!tipc_nametbl_lookup_anycast(net, ua, &skaddr))
1471 return -EHOSTUNREACH;
1472 } else if (atype == TIPC_SOCKET_ADDR) {
1473 skaddr = ua->sk;
1474 } else {
1475 return -EINVAL;
1476 }
1477
1478 /* Block or return if destination link is congested */
1479 rc = tipc_wait_for_cond(sock, &timeout,
1480 !tipc_dest_find(clinks, skaddr.node, 0));
1481 if (unlikely(rc))
1482 return rc;
1483
1484 /* Finally build message header */
1485 msg_set_destnode(hdr, skaddr.node);
1486 msg_set_destport(hdr, skaddr.ref);
1487 if (atype == TIPC_SERVICE_ADDR) {
1488 msg_set_type(hdr, TIPC_NAMED_MSG);
1489 msg_set_hdr_sz(hdr, NAMED_H_SIZE);
1490 msg_set_nametype(hdr, ua->sa.type);
1491 msg_set_nameinst(hdr, ua->sa.instance);
1492 msg_set_lookup_scope(hdr, ua->scope);
1493 } else { /* TIPC_SOCKET_ADDR */
1494 msg_set_type(hdr, TIPC_DIRECT_MSG);
1495 msg_set_lookup_scope(hdr, 0);
1496 msg_set_hdr_sz(hdr, BASIC_H_SIZE);
1497 }
1498
1499 /* Add message body */
1500 __skb_queue_head_init(&pkts);
1501 mtu = tipc_node_get_mtu(net, skaddr.node, tsk->portid, true);
1502 rc = tipc_msg_build(hdr, m, 0, dlen, mtu, &pkts);
1503 if (unlikely(rc != dlen))
1504 return rc;
1505 if (unlikely(syn && !tipc_msg_skb_clone(&pkts, &sk->sk_write_queue))) {
1506 __skb_queue_purge(&pkts);
1507 return -ENOMEM;
1508 }
1509
1510 /* Send message */
1511 trace_tipc_sk_sendmsg(sk, skb_peek(&pkts), TIPC_DUMP_SK_SNDQ, " ");
1512 rc = tipc_node_xmit(net, &pkts, skaddr.node, tsk->portid);
1513 if (unlikely(rc == -ELINKCONG)) {
1514 tipc_dest_push(clinks, skaddr.node, 0);
1515 tsk->cong_link_cnt++;
1516 rc = 0;
1517 }
1518
1519 if (unlikely(syn && !rc)) {
1520 tipc_set_sk_state(sk, TIPC_CONNECTING);
1521 if (dlen && timeout) {
1522 timeout = msecs_to_jiffies(timeout);
1523 tipc_wait_for_connect(sock, &timeout);
1524 }
1525 }
1526
1527 return rc ? rc : dlen;
1528}
1529
1530/**
1531 * tipc_sendstream - send stream-oriented data
1532 * @sock: socket structure
1533 * @m: data to send
1534 * @dsz: total length of data to be transmitted
1535 *
1536 * Used for SOCK_STREAM data.
1537 *
1538 * Return: the number of bytes sent on success (or partial success),
1539 * or errno if no data sent
1540 */
1541static int tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dsz)
1542{
1543 struct sock *sk = sock->sk;
1544 int ret;
1545
1546 lock_sock(sk);
1547 ret = __tipc_sendstream(sock, m, dsz);
1548 release_sock(sk);
1549
1550 return ret;
1551}
1552
1553static int __tipc_sendstream(struct socket *sock, struct msghdr *m, size_t dlen)
1554{
1555 struct sock *sk = sock->sk;
1556 DECLARE_SOCKADDR(struct sockaddr_tipc *, dest, m->msg_name);
1557 long timeout = sock_sndtimeo(sk, m->msg_flags & MSG_DONTWAIT);
1558 struct sk_buff_head *txq = &sk->sk_write_queue;
1559 struct tipc_sock *tsk = tipc_sk(sk);
1560 struct tipc_msg *hdr = &tsk->phdr;
1561 struct net *net = sock_net(sk);
1562 struct sk_buff *skb;
1563 u32 dnode = tsk_peer_node(tsk);
1564 int maxnagle = tsk->maxnagle;
1565 int maxpkt = tsk->max_pkt;
1566 int send, sent = 0;
1567 int blocks, rc = 0;
1568
1569 if (unlikely(dlen > INT_MAX))
1570 return -EMSGSIZE;
1571
1572 /* Handle implicit connection setup */
1573 if (unlikely(dest && sk->sk_state == TIPC_OPEN)) {
1574 rc = __tipc_sendmsg(sock, m, dlen);
1575 if (dlen && dlen == rc) {
1576 tsk->peer_caps = tipc_node_get_capabilities(net, dnode);
1577 tsk->snt_unacked = tsk_inc(tsk, dlen + msg_hdr_sz(hdr));
1578 }
1579 return rc;
1580 }
1581
1582 do {
1583 rc = tipc_wait_for_cond(sock, &timeout,
1584 (!tsk->cong_link_cnt &&
1585 !tsk_conn_cong(tsk) &&
1586 tipc_sk_connected(sk)));
1587 if (unlikely(rc))
1588 break;
1589 send = min_t(size_t, dlen - sent, TIPC_MAX_USER_MSG_SIZE);
1590 blocks = tsk->snd_backlog;
1591 if (tsk->oneway++ >= tsk->nagle_start && maxnagle &&
1592 send <= maxnagle) {
1593 rc = tipc_msg_append(hdr, m, send, maxnagle, txq);
1594 if (unlikely(rc < 0))
1595 break;
1596 blocks += rc;
1597 tsk->msg_acc++;
1598 if (blocks <= 64 && tsk->expect_ack) {
1599 tsk->snd_backlog = blocks;
1600 sent += send;
1601 break;
1602 } else if (blocks > 64) {
1603 tsk->pkt_cnt += skb_queue_len(txq);
1604 } else {
1605 skb = skb_peek_tail(txq);
1606 if (skb) {
1607 msg_set_ack_required(buf_msg(skb));
1608 tsk->expect_ack = true;
1609 } else {
1610 tsk->expect_ack = false;
1611 }
1612 tsk->msg_acc = 0;
1613 tsk->pkt_cnt = 0;
1614 }
1615 } else {
1616 rc = tipc_msg_build(hdr, m, sent, send, maxpkt, txq);
1617 if (unlikely(rc != send))
1618 break;
1619 blocks += tsk_inc(tsk, send + MIN_H_SIZE);
1620 }
1621 trace_tipc_sk_sendstream(sk, skb_peek(txq),
1622 TIPC_DUMP_SK_SNDQ, " ");
1623 rc = tipc_node_xmit(net, txq, dnode, tsk->portid);
1624 if (unlikely(rc == -ELINKCONG)) {
1625 tsk->cong_link_cnt = 1;
1626 rc = 0;
1627 }
1628 if (likely(!rc)) {
1629 tsk->snt_unacked += blocks;
1630 tsk->snd_backlog = 0;
1631 sent += send;
1632 }
1633 } while (sent < dlen && !rc);
1634
1635 return sent ? sent : rc;
1636}
1637
1638/**
1639 * tipc_send_packet - send a connection-oriented message
1640 * @sock: socket structure
1641 * @m: message to send
1642 * @dsz: length of data to be transmitted
1643 *
1644 * Used for SOCK_SEQPACKET messages.
1645 *
1646 * Return: the number of bytes sent on success, or errno otherwise
1647 */
1648static int tipc_send_packet(struct socket *sock, struct msghdr *m, size_t dsz)
1649{
1650 if (dsz > TIPC_MAX_USER_MSG_SIZE)
1651 return -EMSGSIZE;
1652
1653 return tipc_sendstream(sock, m, dsz);
1654}
1655
1656/* tipc_sk_finish_conn - complete the setup of a connection
1657 */
1658static void tipc_sk_finish_conn(struct tipc_sock *tsk, u32 peer_port,
1659 u32 peer_node)
1660{
1661 struct sock *sk = &tsk->sk;
1662 struct net *net = sock_net(sk);
1663 struct tipc_msg *msg = &tsk->phdr;
1664
1665 msg_set_syn(msg, 0);
1666 msg_set_destnode(msg, peer_node);
1667 msg_set_destport(msg, peer_port);
1668 msg_set_type(msg, TIPC_CONN_MSG);
1669 msg_set_lookup_scope(msg, 0);
1670 msg_set_hdr_sz(msg, SHORT_H_SIZE);
1671
1672 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
1673 tipc_set_sk_state(sk, TIPC_ESTABLISHED);
1674 tipc_node_add_conn(net, peer_node, tsk->portid, peer_port);
1675 tsk->max_pkt = tipc_node_get_mtu(net, peer_node, tsk->portid, true);
1676 tsk->peer_caps = tipc_node_get_capabilities(net, peer_node);
1677 tsk_set_nagle(tsk);
1678 __skb_queue_purge(&sk->sk_write_queue);
1679 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL)
1680 return;
1681
1682 /* Fall back to message based flow control */
1683 tsk->rcv_win = FLOWCTL_MSG_WIN;
1684 tsk->snd_win = FLOWCTL_MSG_WIN;
1685}
1686
1687/**
1688 * tipc_sk_set_orig_addr - capture sender's address for received message
1689 * @m: descriptor for message info
1690 * @skb: received message
1691 *
1692 * Note: Address is not captured if not requested by receiver.
1693 */
1694static void tipc_sk_set_orig_addr(struct msghdr *m, struct sk_buff *skb)
1695{
1696 DECLARE_SOCKADDR(struct sockaddr_pair *, srcaddr, m->msg_name);
1697 struct tipc_msg *hdr = buf_msg(skb);
1698
1699 if (!srcaddr)
1700 return;
1701
1702 srcaddr->sock.family = AF_TIPC;
1703 srcaddr->sock.addrtype = TIPC_SOCKET_ADDR;
1704 srcaddr->sock.scope = 0;
1705 srcaddr->sock.addr.id.ref = msg_origport(hdr);
1706 srcaddr->sock.addr.id.node = msg_orignode(hdr);
1707 srcaddr->sock.addr.name.domain = 0;
1708 m->msg_namelen = sizeof(struct sockaddr_tipc);
1709
1710 if (!msg_in_group(hdr))
1711 return;
1712
1713 /* Group message users may also want to know sending member's id */
1714 srcaddr->member.family = AF_TIPC;
1715 srcaddr->member.addrtype = TIPC_SERVICE_ADDR;
1716 srcaddr->member.scope = 0;
1717 srcaddr->member.addr.name.name.type = msg_nametype(hdr);
1718 srcaddr->member.addr.name.name.instance = TIPC_SKB_CB(skb)->orig_member;
1719 srcaddr->member.addr.name.domain = 0;
1720 m->msg_namelen = sizeof(*srcaddr);
1721}
1722
1723/**
1724 * tipc_sk_anc_data_recv - optionally capture ancillary data for received message
1725 * @m: descriptor for message info
1726 * @skb: received message buffer
1727 * @tsk: TIPC port associated with message
1728 *
1729 * Note: Ancillary data is not captured if not requested by receiver.
1730 *
1731 * Return: 0 if successful, otherwise errno
1732 */
1733static int tipc_sk_anc_data_recv(struct msghdr *m, struct sk_buff *skb,
1734 struct tipc_sock *tsk)
1735{
1736 struct tipc_msg *hdr;
1737 u32 data[3] = {0,};
1738 bool has_addr;
1739 int dlen, rc;
1740
1741 if (likely(m->msg_controllen == 0))
1742 return 0;
1743
1744 hdr = buf_msg(skb);
1745 dlen = msg_data_sz(hdr);
1746
1747 /* Capture errored message object, if any */
1748 if (msg_errcode(hdr)) {
1749 if (skb_linearize(skb))
1750 return -ENOMEM;
1751 hdr = buf_msg(skb);
1752 data[0] = msg_errcode(hdr);
1753 data[1] = dlen;
1754 rc = put_cmsg(m, SOL_TIPC, TIPC_ERRINFO, 8, data);
1755 if (rc || !dlen)
1756 return rc;
1757 rc = put_cmsg(m, SOL_TIPC, TIPC_RETDATA, dlen, msg_data(hdr));
1758 if (rc)
1759 return rc;
1760 }
1761
1762 /* Capture TIPC_SERVICE_ADDR/RANGE destination address, if any */
1763 switch (msg_type(hdr)) {
1764 case TIPC_NAMED_MSG:
1765 has_addr = true;
1766 data[0] = msg_nametype(hdr);
1767 data[1] = msg_namelower(hdr);
1768 data[2] = data[1];
1769 break;
1770 case TIPC_MCAST_MSG:
1771 has_addr = true;
1772 data[0] = msg_nametype(hdr);
1773 data[1] = msg_namelower(hdr);
1774 data[2] = msg_nameupper(hdr);
1775 break;
1776 case TIPC_CONN_MSG:
1777 has_addr = !!tsk->conn_addrtype;
1778 data[0] = msg_nametype(&tsk->phdr);
1779 data[1] = msg_nameinst(&tsk->phdr);
1780 data[2] = data[1];
1781 break;
1782 default:
1783 has_addr = false;
1784 }
1785 if (!has_addr)
1786 return 0;
1787 return put_cmsg(m, SOL_TIPC, TIPC_DESTNAME, 12, data);
1788}
1789
1790static struct sk_buff *tipc_sk_build_ack(struct tipc_sock *tsk)
1791{
1792 struct sock *sk = &tsk->sk;
1793 struct sk_buff *skb = NULL;
1794 struct tipc_msg *msg;
1795 u32 peer_port = tsk_peer_port(tsk);
1796 u32 dnode = tsk_peer_node(tsk);
1797
1798 if (!tipc_sk_connected(sk))
1799 return NULL;
1800 skb = tipc_msg_create(CONN_MANAGER, CONN_ACK, INT_H_SIZE, 0,
1801 dnode, tsk_own_node(tsk), peer_port,
1802 tsk->portid, TIPC_OK);
1803 if (!skb)
1804 return NULL;
1805 msg = buf_msg(skb);
1806 msg_set_conn_ack(msg, tsk->rcv_unacked);
1807 tsk->rcv_unacked = 0;
1808
1809 /* Adjust to and advertize the correct window limit */
1810 if (tsk->peer_caps & TIPC_BLOCK_FLOWCTL) {
1811 tsk->rcv_win = tsk_adv_blocks(tsk->sk.sk_rcvbuf);
1812 msg_set_adv_win(msg, tsk->rcv_win);
1813 }
1814 return skb;
1815}
1816
1817static void tipc_sk_send_ack(struct tipc_sock *tsk)
1818{
1819 struct sk_buff *skb;
1820
1821 skb = tipc_sk_build_ack(tsk);
1822 if (!skb)
1823 return;
1824
1825 tipc_node_xmit_skb(sock_net(&tsk->sk), skb, tsk_peer_node(tsk),
1826 msg_link_selector(buf_msg(skb)));
1827}
1828
1829static int tipc_wait_for_rcvmsg(struct socket *sock, long *timeop)
1830{
1831 struct sock *sk = sock->sk;
1832 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1833 long timeo = *timeop;
1834 int err = sock_error(sk);
1835
1836 if (err)
1837 return err;
1838
1839 for (;;) {
1840 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
1841 if (sk->sk_shutdown & RCV_SHUTDOWN) {
1842 err = -ENOTCONN;
1843 break;
1844 }
1845 add_wait_queue(sk_sleep(sk), &wait);
1846 release_sock(sk);
1847 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
1848 sched_annotate_sleep();
1849 lock_sock(sk);
1850 remove_wait_queue(sk_sleep(sk), &wait);
1851 }
1852 err = 0;
1853 if (!skb_queue_empty(&sk->sk_receive_queue))
1854 break;
1855 err = -EAGAIN;
1856 if (!timeo)
1857 break;
1858 err = sock_intr_errno(timeo);
1859 if (signal_pending(current))
1860 break;
1861
1862 err = sock_error(sk);
1863 if (err)
1864 break;
1865 }
1866 *timeop = timeo;
1867 return err;
1868}
1869
1870/**
1871 * tipc_recvmsg - receive packet-oriented message
1872 * @sock: network socket
1873 * @m: descriptor for message info
1874 * @buflen: length of user buffer area
1875 * @flags: receive flags
1876 *
1877 * Used for SOCK_DGRAM, SOCK_RDM, and SOCK_SEQPACKET messages.
1878 * If the complete message doesn't fit in user area, truncate it.
1879 *
1880 * Return: size of returned message data, errno otherwise
1881 */
1882static int tipc_recvmsg(struct socket *sock, struct msghdr *m,
1883 size_t buflen, int flags)
1884{
1885 struct sock *sk = sock->sk;
1886 bool connected = !tipc_sk_type_connectionless(sk);
1887 struct tipc_sock *tsk = tipc_sk(sk);
1888 int rc, err, hlen, dlen, copy;
1889 struct tipc_skb_cb *skb_cb;
1890 struct sk_buff_head xmitq;
1891 struct tipc_msg *hdr;
1892 struct sk_buff *skb;
1893 bool grp_evt;
1894 long timeout;
1895
1896 /* Catch invalid receive requests */
1897 if (unlikely(!buflen))
1898 return -EINVAL;
1899
1900 lock_sock(sk);
1901 if (unlikely(connected && sk->sk_state == TIPC_OPEN)) {
1902 rc = -ENOTCONN;
1903 goto exit;
1904 }
1905 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
1906
1907 /* Step rcv queue to first msg with data or error; wait if necessary */
1908 do {
1909 rc = tipc_wait_for_rcvmsg(sock, &timeout);
1910 if (unlikely(rc))
1911 goto exit;
1912 skb = skb_peek(&sk->sk_receive_queue);
1913 skb_cb = TIPC_SKB_CB(skb);
1914 hdr = buf_msg(skb);
1915 dlen = msg_data_sz(hdr);
1916 hlen = msg_hdr_sz(hdr);
1917 err = msg_errcode(hdr);
1918 grp_evt = msg_is_grp_evt(hdr);
1919 if (likely(dlen || err))
1920 break;
1921 tsk_advance_rx_queue(sk);
1922 } while (1);
1923
1924 /* Collect msg meta data, including error code and rejected data */
1925 tipc_sk_set_orig_addr(m, skb);
1926 rc = tipc_sk_anc_data_recv(m, skb, tsk);
1927 if (unlikely(rc))
1928 goto exit;
1929 hdr = buf_msg(skb);
1930
1931 /* Capture data if non-error msg, otherwise just set return value */
1932 if (likely(!err)) {
1933 int offset = skb_cb->bytes_read;
1934
1935 copy = min_t(int, dlen - offset, buflen);
1936 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
1937 if (unlikely(rc))
1938 goto exit;
1939 if (unlikely(offset + copy < dlen)) {
1940 if (flags & MSG_EOR) {
1941 if (!(flags & MSG_PEEK))
1942 skb_cb->bytes_read = offset + copy;
1943 } else {
1944 m->msg_flags |= MSG_TRUNC;
1945 skb_cb->bytes_read = 0;
1946 }
1947 } else {
1948 if (flags & MSG_EOR)
1949 m->msg_flags |= MSG_EOR;
1950 skb_cb->bytes_read = 0;
1951 }
1952 } else {
1953 copy = 0;
1954 rc = 0;
1955 if (err != TIPC_CONN_SHUTDOWN && connected && !m->msg_control) {
1956 rc = -ECONNRESET;
1957 goto exit;
1958 }
1959 }
1960
1961 /* Mark message as group event if applicable */
1962 if (unlikely(grp_evt)) {
1963 if (msg_grp_evt(hdr) == TIPC_WITHDRAWN)
1964 m->msg_flags |= MSG_EOR;
1965 m->msg_flags |= MSG_OOB;
1966 copy = 0;
1967 }
1968
1969 /* Caption of data or error code/rejected data was successful */
1970 if (unlikely(flags & MSG_PEEK))
1971 goto exit;
1972
1973 /* Send group flow control advertisement when applicable */
1974 if (tsk->group && msg_in_group(hdr) && !grp_evt) {
1975 __skb_queue_head_init(&xmitq);
1976 tipc_group_update_rcv_win(tsk->group, tsk_blocks(hlen + dlen),
1977 msg_orignode(hdr), msg_origport(hdr),
1978 &xmitq);
1979 tipc_node_distr_xmit(sock_net(sk), &xmitq);
1980 }
1981
1982 if (skb_cb->bytes_read)
1983 goto exit;
1984
1985 tsk_advance_rx_queue(sk);
1986
1987 if (likely(!connected))
1988 goto exit;
1989
1990 /* Send connection flow control advertisement when applicable */
1991 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
1992 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
1993 tipc_sk_send_ack(tsk);
1994exit:
1995 release_sock(sk);
1996 return rc ? rc : copy;
1997}
1998
1999/**
2000 * tipc_recvstream - receive stream-oriented data
2001 * @sock: network socket
2002 * @m: descriptor for message info
2003 * @buflen: total size of user buffer area
2004 * @flags: receive flags
2005 *
2006 * Used for SOCK_STREAM messages only. If not enough data is available
2007 * will optionally wait for more; never truncates data.
2008 *
2009 * Return: size of returned message data, errno otherwise
2010 */
2011static int tipc_recvstream(struct socket *sock, struct msghdr *m,
2012 size_t buflen, int flags)
2013{
2014 struct sock *sk = sock->sk;
2015 struct tipc_sock *tsk = tipc_sk(sk);
2016 struct sk_buff *skb;
2017 struct tipc_msg *hdr;
2018 struct tipc_skb_cb *skb_cb;
2019 bool peek = flags & MSG_PEEK;
2020 int offset, required, copy, copied = 0;
2021 int hlen, dlen, err, rc;
2022 long timeout;
2023
2024 /* Catch invalid receive attempts */
2025 if (unlikely(!buflen))
2026 return -EINVAL;
2027
2028 lock_sock(sk);
2029
2030 if (unlikely(sk->sk_state == TIPC_OPEN)) {
2031 rc = -ENOTCONN;
2032 goto exit;
2033 }
2034 required = sock_rcvlowat(sk, flags & MSG_WAITALL, buflen);
2035 timeout = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
2036
2037 do {
2038 /* Look at first msg in receive queue; wait if necessary */
2039 rc = tipc_wait_for_rcvmsg(sock, &timeout);
2040 if (unlikely(rc))
2041 break;
2042 skb = skb_peek(&sk->sk_receive_queue);
2043 skb_cb = TIPC_SKB_CB(skb);
2044 hdr = buf_msg(skb);
2045 dlen = msg_data_sz(hdr);
2046 hlen = msg_hdr_sz(hdr);
2047 err = msg_errcode(hdr);
2048
2049 /* Discard any empty non-errored (SYN-) message */
2050 if (unlikely(!dlen && !err)) {
2051 tsk_advance_rx_queue(sk);
2052 continue;
2053 }
2054
2055 /* Collect msg meta data, incl. error code and rejected data */
2056 if (!copied) {
2057 tipc_sk_set_orig_addr(m, skb);
2058 rc = tipc_sk_anc_data_recv(m, skb, tsk);
2059 if (rc)
2060 break;
2061 hdr = buf_msg(skb);
2062 }
2063
2064 /* Copy data if msg ok, otherwise return error/partial data */
2065 if (likely(!err)) {
2066 offset = skb_cb->bytes_read;
2067 copy = min_t(int, dlen - offset, buflen - copied);
2068 rc = skb_copy_datagram_msg(skb, hlen + offset, m, copy);
2069 if (unlikely(rc))
2070 break;
2071 copied += copy;
2072 offset += copy;
2073 if (unlikely(offset < dlen)) {
2074 if (!peek)
2075 skb_cb->bytes_read = offset;
2076 break;
2077 }
2078 } else {
2079 rc = 0;
2080 if ((err != TIPC_CONN_SHUTDOWN) && !m->msg_control)
2081 rc = -ECONNRESET;
2082 if (copied || rc)
2083 break;
2084 }
2085
2086 if (unlikely(peek))
2087 break;
2088
2089 tsk_advance_rx_queue(sk);
2090
2091 /* Send connection flow control advertisement when applicable */
2092 tsk->rcv_unacked += tsk_inc(tsk, hlen + dlen);
2093 if (tsk->rcv_unacked >= tsk->rcv_win / TIPC_ACK_RATE)
2094 tipc_sk_send_ack(tsk);
2095
2096 /* Exit if all requested data or FIN/error received */
2097 if (copied == buflen || err)
2098 break;
2099
2100 } while (!skb_queue_empty(&sk->sk_receive_queue) || copied < required);
2101exit:
2102 release_sock(sk);
2103 return copied ? copied : rc;
2104}
2105
2106/**
2107 * tipc_write_space - wake up thread if port congestion is released
2108 * @sk: socket
2109 */
2110static void tipc_write_space(struct sock *sk)
2111{
2112 struct socket_wq *wq;
2113
2114 rcu_read_lock();
2115 wq = rcu_dereference(sk->sk_wq);
2116 if (skwq_has_sleeper(wq))
2117 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2118 EPOLLWRNORM | EPOLLWRBAND);
2119 rcu_read_unlock();
2120}
2121
2122/**
2123 * tipc_data_ready - wake up threads to indicate messages have been received
2124 * @sk: socket
2125 */
2126static void tipc_data_ready(struct sock *sk)
2127{
2128 struct socket_wq *wq;
2129
2130 rcu_read_lock();
2131 wq = rcu_dereference(sk->sk_wq);
2132 if (skwq_has_sleeper(wq))
2133 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN |
2134 EPOLLRDNORM | EPOLLRDBAND);
2135 rcu_read_unlock();
2136}
2137
2138static void tipc_sock_destruct(struct sock *sk)
2139{
2140 __skb_queue_purge(&sk->sk_receive_queue);
2141}
2142
2143static void tipc_sk_proto_rcv(struct sock *sk,
2144 struct sk_buff_head *inputq,
2145 struct sk_buff_head *xmitq)
2146{
2147 struct sk_buff *skb = __skb_dequeue(inputq);
2148 struct tipc_sock *tsk = tipc_sk(sk);
2149 struct tipc_msg *hdr = buf_msg(skb);
2150 struct tipc_group *grp = tsk->group;
2151 bool wakeup = false;
2152
2153 switch (msg_user(hdr)) {
2154 case CONN_MANAGER:
2155 tipc_sk_conn_proto_rcv(tsk, skb, inputq, xmitq);
2156 return;
2157 case SOCK_WAKEUP:
2158 tipc_dest_del(&tsk->cong_links, msg_orignode(hdr), 0);
2159 /* coupled with smp_rmb() in tipc_wait_for_cond() */
2160 smp_wmb();
2161 tsk->cong_link_cnt--;
2162 wakeup = true;
2163 tipc_sk_push_backlog(tsk, false);
2164 break;
2165 case GROUP_PROTOCOL:
2166 tipc_group_proto_rcv(grp, &wakeup, hdr, inputq, xmitq);
2167 break;
2168 case TOP_SRV:
2169 tipc_group_member_evt(tsk->group, &wakeup, &sk->sk_rcvbuf,
2170 hdr, inputq, xmitq);
2171 break;
2172 default:
2173 break;
2174 }
2175
2176 if (wakeup)
2177 sk->sk_write_space(sk);
2178
2179 kfree_skb(skb);
2180}
2181
2182/**
2183 * tipc_sk_filter_connect - check incoming message for a connection-based socket
2184 * @tsk: TIPC socket
2185 * @skb: pointer to message buffer.
2186 * @xmitq: for Nagle ACK if any
2187 * Return: true if message should be added to receive queue, false otherwise
2188 */
2189static bool tipc_sk_filter_connect(struct tipc_sock *tsk, struct sk_buff *skb,
2190 struct sk_buff_head *xmitq)
2191{
2192 struct sock *sk = &tsk->sk;
2193 struct net *net = sock_net(sk);
2194 struct tipc_msg *hdr = buf_msg(skb);
2195 bool con_msg = msg_connected(hdr);
2196 u32 pport = tsk_peer_port(tsk);
2197 u32 pnode = tsk_peer_node(tsk);
2198 u32 oport = msg_origport(hdr);
2199 u32 onode = msg_orignode(hdr);
2200 int err = msg_errcode(hdr);
2201 unsigned long delay;
2202
2203 if (unlikely(msg_mcast(hdr)))
2204 return false;
2205 tsk->oneway = 0;
2206
2207 switch (sk->sk_state) {
2208 case TIPC_CONNECTING:
2209 /* Setup ACK */
2210 if (likely(con_msg)) {
2211 if (err)
2212 break;
2213 tipc_sk_finish_conn(tsk, oport, onode);
2214 msg_set_importance(&tsk->phdr, msg_importance(hdr));
2215 /* ACK+ message with data is added to receive queue */
2216 if (msg_data_sz(hdr))
2217 return true;
2218 /* Empty ACK-, - wake up sleeping connect() and drop */
2219 sk->sk_state_change(sk);
2220 msg_set_dest_droppable(hdr, 1);
2221 return false;
2222 }
2223 /* Ignore connectionless message if not from listening socket */
2224 if (oport != pport || onode != pnode)
2225 return false;
2226
2227 /* Rejected SYN */
2228 if (err != TIPC_ERR_OVERLOAD)
2229 break;
2230
2231 /* Prepare for new setup attempt if we have a SYN clone */
2232 if (skb_queue_empty(&sk->sk_write_queue))
2233 break;
2234 get_random_bytes(&delay, 2);
2235 delay %= (tsk->conn_timeout / 4);
2236 delay = msecs_to_jiffies(delay + 100);
2237 sk_reset_timer(sk, &sk->sk_timer, jiffies + delay);
2238 return false;
2239 case TIPC_OPEN:
2240 case TIPC_DISCONNECTING:
2241 return false;
2242 case TIPC_LISTEN:
2243 /* Accept only SYN message */
2244 if (!msg_is_syn(hdr) &&
2245 tipc_node_get_capabilities(net, onode) & TIPC_SYN_BIT)
2246 return false;
2247 if (!con_msg && !err)
2248 return true;
2249 return false;
2250 case TIPC_ESTABLISHED:
2251 if (!skb_queue_empty(&sk->sk_write_queue))
2252 tipc_sk_push_backlog(tsk, false);
2253 /* Accept only connection-based messages sent by peer */
2254 if (likely(con_msg && !err && pport == oport &&
2255 pnode == onode)) {
2256 if (msg_ack_required(hdr)) {
2257 struct sk_buff *skb;
2258
2259 skb = tipc_sk_build_ack(tsk);
2260 if (skb) {
2261 msg_set_nagle_ack(buf_msg(skb));
2262 __skb_queue_tail(xmitq, skb);
2263 }
2264 }
2265 return true;
2266 }
2267 if (!tsk_peer_msg(tsk, hdr))
2268 return false;
2269 if (!err)
2270 return true;
2271 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2272 tipc_node_remove_conn(net, pnode, tsk->portid);
2273 sk->sk_state_change(sk);
2274 return true;
2275 default:
2276 pr_err("Unknown sk_state %u\n", sk->sk_state);
2277 }
2278 /* Abort connection setup attempt */
2279 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2280 sk->sk_err = ECONNREFUSED;
2281 sk->sk_state_change(sk);
2282 return true;
2283}
2284
2285/**
2286 * rcvbuf_limit - get proper overload limit of socket receive queue
2287 * @sk: socket
2288 * @skb: message
2289 *
2290 * For connection oriented messages, irrespective of importance,
2291 * default queue limit is 2 MB.
2292 *
2293 * For connectionless messages, queue limits are based on message
2294 * importance as follows:
2295 *
2296 * TIPC_LOW_IMPORTANCE (2 MB)
2297 * TIPC_MEDIUM_IMPORTANCE (4 MB)
2298 * TIPC_HIGH_IMPORTANCE (8 MB)
2299 * TIPC_CRITICAL_IMPORTANCE (16 MB)
2300 *
2301 * Return: overload limit according to corresponding message importance
2302 */
2303static unsigned int rcvbuf_limit(struct sock *sk, struct sk_buff *skb)
2304{
2305 struct tipc_sock *tsk = tipc_sk(sk);
2306 struct tipc_msg *hdr = buf_msg(skb);
2307
2308 if (unlikely(msg_in_group(hdr)))
2309 return READ_ONCE(sk->sk_rcvbuf);
2310
2311 if (unlikely(!msg_connected(hdr)))
2312 return READ_ONCE(sk->sk_rcvbuf) << msg_importance(hdr);
2313
2314 if (likely(tsk->peer_caps & TIPC_BLOCK_FLOWCTL))
2315 return READ_ONCE(sk->sk_rcvbuf);
2316
2317 return FLOWCTL_MSG_LIM;
2318}
2319
2320/**
2321 * tipc_sk_filter_rcv - validate incoming message
2322 * @sk: socket
2323 * @skb: pointer to message.
2324 * @xmitq: output message area (FIXME)
2325 *
2326 * Enqueues message on receive queue if acceptable; optionally handles
2327 * disconnect indication for a connected socket.
2328 *
2329 * Called with socket lock already taken
2330 */
2331static void tipc_sk_filter_rcv(struct sock *sk, struct sk_buff *skb,
2332 struct sk_buff_head *xmitq)
2333{
2334 bool sk_conn = !tipc_sk_type_connectionless(sk);
2335 struct tipc_sock *tsk = tipc_sk(sk);
2336 struct tipc_group *grp = tsk->group;
2337 struct tipc_msg *hdr = buf_msg(skb);
2338 struct net *net = sock_net(sk);
2339 struct sk_buff_head inputq;
2340 int mtyp = msg_type(hdr);
2341 int limit, err = TIPC_OK;
2342
2343 trace_tipc_sk_filter_rcv(sk, skb, TIPC_DUMP_ALL, " ");
2344 TIPC_SKB_CB(skb)->bytes_read = 0;
2345 __skb_queue_head_init(&inputq);
2346 __skb_queue_tail(&inputq, skb);
2347
2348 if (unlikely(!msg_isdata(hdr)))
2349 tipc_sk_proto_rcv(sk, &inputq, xmitq);
2350
2351 if (unlikely(grp))
2352 tipc_group_filter_msg(grp, &inputq, xmitq);
2353
2354 if (unlikely(!grp) && mtyp == TIPC_MCAST_MSG)
2355 tipc_mcast_filter_msg(net, &tsk->mc_method.deferredq, &inputq);
2356
2357 /* Validate and add to receive buffer if there is space */
2358 while ((skb = __skb_dequeue(&inputq))) {
2359 hdr = buf_msg(skb);
2360 limit = rcvbuf_limit(sk, skb);
2361 if ((sk_conn && !tipc_sk_filter_connect(tsk, skb, xmitq)) ||
2362 (!sk_conn && msg_connected(hdr)) ||
2363 (!grp && msg_in_group(hdr)))
2364 err = TIPC_ERR_NO_PORT;
2365 else if (sk_rmem_alloc_get(sk) + skb->truesize >= limit) {
2366 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL,
2367 "err_overload2!");
2368 atomic_inc(&sk->sk_drops);
2369 err = TIPC_ERR_OVERLOAD;
2370 }
2371
2372 if (unlikely(err)) {
2373 if (tipc_msg_reverse(tipc_own_addr(net), &skb, err)) {
2374 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_NONE,
2375 "@filter_rcv!");
2376 __skb_queue_tail(xmitq, skb);
2377 }
2378 err = TIPC_OK;
2379 continue;
2380 }
2381 __skb_queue_tail(&sk->sk_receive_queue, skb);
2382 skb_set_owner_r(skb, sk);
2383 trace_tipc_sk_overlimit2(sk, skb, TIPC_DUMP_ALL,
2384 "rcvq >90% allocated!");
2385 sk->sk_data_ready(sk);
2386 }
2387}
2388
2389/**
2390 * tipc_sk_backlog_rcv - handle incoming message from backlog queue
2391 * @sk: socket
2392 * @skb: message
2393 *
2394 * Caller must hold socket lock
2395 */
2396static int tipc_sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
2397{
2398 unsigned int before = sk_rmem_alloc_get(sk);
2399 struct sk_buff_head xmitq;
2400 unsigned int added;
2401
2402 __skb_queue_head_init(&xmitq);
2403
2404 tipc_sk_filter_rcv(sk, skb, &xmitq);
2405 added = sk_rmem_alloc_get(sk) - before;
2406 atomic_add(added, &tipc_sk(sk)->dupl_rcvcnt);
2407
2408 /* Send pending response/rejected messages, if any */
2409 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2410 return 0;
2411}
2412
2413/**
2414 * tipc_sk_enqueue - extract all buffers with destination 'dport' from
2415 * inputq and try adding them to socket or backlog queue
2416 * @inputq: list of incoming buffers with potentially different destinations
2417 * @sk: socket where the buffers should be enqueued
2418 * @dport: port number for the socket
2419 * @xmitq: output queue
2420 *
2421 * Caller must hold socket lock
2422 */
2423static void tipc_sk_enqueue(struct sk_buff_head *inputq, struct sock *sk,
2424 u32 dport, struct sk_buff_head *xmitq)
2425{
2426 unsigned long time_limit = jiffies + usecs_to_jiffies(20000);
2427 struct sk_buff *skb;
2428 unsigned int lim;
2429 atomic_t *dcnt;
2430 u32 onode;
2431
2432 while (skb_queue_len(inputq)) {
2433 if (unlikely(time_after_eq(jiffies, time_limit)))
2434 return;
2435
2436 skb = tipc_skb_dequeue(inputq, dport);
2437 if (unlikely(!skb))
2438 return;
2439
2440 /* Add message directly to receive queue if possible */
2441 if (!sock_owned_by_user(sk)) {
2442 tipc_sk_filter_rcv(sk, skb, xmitq);
2443 continue;
2444 }
2445
2446 /* Try backlog, compensating for double-counted bytes */
2447 dcnt = &tipc_sk(sk)->dupl_rcvcnt;
2448 if (!sk->sk_backlog.len)
2449 atomic_set(dcnt, 0);
2450 lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
2451 if (likely(!sk_add_backlog(sk, skb, lim))) {
2452 trace_tipc_sk_overlimit1(sk, skb, TIPC_DUMP_ALL,
2453 "bklg & rcvq >90% allocated!");
2454 continue;
2455 }
2456
2457 trace_tipc_sk_dump(sk, skb, TIPC_DUMP_ALL, "err_overload!");
2458 /* Overload => reject message back to sender */
2459 onode = tipc_own_addr(sock_net(sk));
2460 atomic_inc(&sk->sk_drops);
2461 if (tipc_msg_reverse(onode, &skb, TIPC_ERR_OVERLOAD)) {
2462 trace_tipc_sk_rej_msg(sk, skb, TIPC_DUMP_ALL,
2463 "@sk_enqueue!");
2464 __skb_queue_tail(xmitq, skb);
2465 }
2466 break;
2467 }
2468}
2469
2470/**
2471 * tipc_sk_rcv - handle a chain of incoming buffers
2472 * @net: the associated network namespace
2473 * @inputq: buffer list containing the buffers
2474 * Consumes all buffers in list until inputq is empty
2475 * Note: may be called in multiple threads referring to the same queue
2476 */
2477void tipc_sk_rcv(struct net *net, struct sk_buff_head *inputq)
2478{
2479 struct sk_buff_head xmitq;
2480 u32 dnode, dport = 0;
2481 int err;
2482 struct tipc_sock *tsk;
2483 struct sock *sk;
2484 struct sk_buff *skb;
2485
2486 __skb_queue_head_init(&xmitq);
2487 while (skb_queue_len(inputq)) {
2488 dport = tipc_skb_peek_port(inputq, dport);
2489 tsk = tipc_sk_lookup(net, dport);
2490
2491 if (likely(tsk)) {
2492 sk = &tsk->sk;
2493 if (likely(spin_trylock_bh(&sk->sk_lock.slock))) {
2494 tipc_sk_enqueue(inputq, sk, dport, &xmitq);
2495 spin_unlock_bh(&sk->sk_lock.slock);
2496 }
2497 /* Send pending response/rejected messages, if any */
2498 tipc_node_distr_xmit(sock_net(sk), &xmitq);
2499 sock_put(sk);
2500 continue;
2501 }
2502 /* No destination socket => dequeue skb if still there */
2503 skb = tipc_skb_dequeue(inputq, dport);
2504 if (!skb)
2505 return;
2506
2507 /* Try secondary lookup if unresolved named message */
2508 err = TIPC_ERR_NO_PORT;
2509 if (tipc_msg_lookup_dest(net, skb, &err))
2510 goto xmit;
2511
2512 /* Prepare for message rejection */
2513 if (!tipc_msg_reverse(tipc_own_addr(net), &skb, err))
2514 continue;
2515
2516 trace_tipc_sk_rej_msg(NULL, skb, TIPC_DUMP_NONE, "@sk_rcv!");
2517xmit:
2518 dnode = msg_destnode(buf_msg(skb));
2519 tipc_node_xmit_skb(net, skb, dnode, dport);
2520 }
2521}
2522
2523static int tipc_wait_for_connect(struct socket *sock, long *timeo_p)
2524{
2525 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2526 struct sock *sk = sock->sk;
2527 int done;
2528
2529 do {
2530 int err = sock_error(sk);
2531 if (err)
2532 return err;
2533 if (!*timeo_p)
2534 return -ETIMEDOUT;
2535 if (signal_pending(current))
2536 return sock_intr_errno(*timeo_p);
2537 if (sk->sk_state == TIPC_DISCONNECTING)
2538 break;
2539
2540 add_wait_queue(sk_sleep(sk), &wait);
2541 done = sk_wait_event(sk, timeo_p, tipc_sk_connected(sk),
2542 &wait);
2543 remove_wait_queue(sk_sleep(sk), &wait);
2544 } while (!done);
2545 return 0;
2546}
2547
2548static bool tipc_sockaddr_is_sane(struct sockaddr_tipc *addr)
2549{
2550 if (addr->family != AF_TIPC)
2551 return false;
2552 if (addr->addrtype == TIPC_SERVICE_RANGE)
2553 return (addr->addr.nameseq.lower <= addr->addr.nameseq.upper);
2554 return (addr->addrtype == TIPC_SERVICE_ADDR ||
2555 addr->addrtype == TIPC_SOCKET_ADDR);
2556}
2557
2558/**
2559 * tipc_connect - establish a connection to another TIPC port
2560 * @sock: socket structure
2561 * @dest: socket address for destination port
2562 * @destlen: size of socket address data structure
2563 * @flags: file-related flags associated with socket
2564 *
2565 * Return: 0 on success, errno otherwise
2566 */
2567static int tipc_connect(struct socket *sock, struct sockaddr *dest,
2568 int destlen, int flags)
2569{
2570 struct sock *sk = sock->sk;
2571 struct tipc_sock *tsk = tipc_sk(sk);
2572 struct sockaddr_tipc *dst = (struct sockaddr_tipc *)dest;
2573 struct msghdr m = {NULL,};
2574 long timeout = (flags & O_NONBLOCK) ? 0 : tsk->conn_timeout;
2575 int previous;
2576 int res = 0;
2577
2578 if (destlen != sizeof(struct sockaddr_tipc))
2579 return -EINVAL;
2580
2581 lock_sock(sk);
2582
2583 if (tsk->group) {
2584 res = -EINVAL;
2585 goto exit;
2586 }
2587
2588 if (dst->family == AF_UNSPEC) {
2589 memset(&tsk->peer, 0, sizeof(struct sockaddr_tipc));
2590 if (!tipc_sk_type_connectionless(sk))
2591 res = -EINVAL;
2592 goto exit;
2593 }
2594 if (!tipc_sockaddr_is_sane(dst)) {
2595 res = -EINVAL;
2596 goto exit;
2597 }
2598 /* DGRAM/RDM connect(), just save the destaddr */
2599 if (tipc_sk_type_connectionless(sk)) {
2600 memcpy(&tsk->peer, dest, destlen);
2601 goto exit;
2602 } else if (dst->addrtype == TIPC_SERVICE_RANGE) {
2603 res = -EINVAL;
2604 goto exit;
2605 }
2606
2607 previous = sk->sk_state;
2608
2609 switch (sk->sk_state) {
2610 case TIPC_OPEN:
2611 /* Send a 'SYN-' to destination */
2612 m.msg_name = dest;
2613 m.msg_namelen = destlen;
2614
2615 /* If connect is in non-blocking case, set MSG_DONTWAIT to
2616 * indicate send_msg() is never blocked.
2617 */
2618 if (!timeout)
2619 m.msg_flags = MSG_DONTWAIT;
2620
2621 res = __tipc_sendmsg(sock, &m, 0);
2622 if ((res < 0) && (res != -EWOULDBLOCK))
2623 goto exit;
2624
2625 /* Just entered TIPC_CONNECTING state; the only
2626 * difference is that return value in non-blocking
2627 * case is EINPROGRESS, rather than EALREADY.
2628 */
2629 res = -EINPROGRESS;
2630 fallthrough;
2631 case TIPC_CONNECTING:
2632 if (!timeout) {
2633 if (previous == TIPC_CONNECTING)
2634 res = -EALREADY;
2635 goto exit;
2636 }
2637 timeout = msecs_to_jiffies(timeout);
2638 /* Wait until an 'ACK' or 'RST' arrives, or a timeout occurs */
2639 res = tipc_wait_for_connect(sock, &timeout);
2640 break;
2641 case TIPC_ESTABLISHED:
2642 res = -EISCONN;
2643 break;
2644 default:
2645 res = -EINVAL;
2646 }
2647
2648exit:
2649 release_sock(sk);
2650 return res;
2651}
2652
2653/**
2654 * tipc_listen - allow socket to listen for incoming connections
2655 * @sock: socket structure
2656 * @len: (unused)
2657 *
2658 * Return: 0 on success, errno otherwise
2659 */
2660static int tipc_listen(struct socket *sock, int len)
2661{
2662 struct sock *sk = sock->sk;
2663 int res;
2664
2665 lock_sock(sk);
2666 res = tipc_set_sk_state(sk, TIPC_LISTEN);
2667 release_sock(sk);
2668
2669 return res;
2670}
2671
2672static int tipc_wait_for_accept(struct socket *sock, long timeo)
2673{
2674 struct sock *sk = sock->sk;
2675 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2676 int err;
2677
2678 /* True wake-one mechanism for incoming connections: only
2679 * one process gets woken up, not the 'whole herd'.
2680 * Since we do not 'race & poll' for established sockets
2681 * anymore, the common case will execute the loop only once.
2682 */
2683 for (;;) {
2684 if (timeo && skb_queue_empty(&sk->sk_receive_queue)) {
2685 add_wait_queue(sk_sleep(sk), &wait);
2686 release_sock(sk);
2687 timeo = wait_woken(&wait, TASK_INTERRUPTIBLE, timeo);
2688 lock_sock(sk);
2689 remove_wait_queue(sk_sleep(sk), &wait);
2690 }
2691 err = 0;
2692 if (!skb_queue_empty(&sk->sk_receive_queue))
2693 break;
2694 err = -EAGAIN;
2695 if (!timeo)
2696 break;
2697 err = sock_intr_errno(timeo);
2698 if (signal_pending(current))
2699 break;
2700 }
2701 return err;
2702}
2703
2704/**
2705 * tipc_accept - wait for connection request
2706 * @sock: listening socket
2707 * @new_sock: new socket that is to be connected
2708 * @flags: file-related flags associated with socket
2709 * @kern: caused by kernel or by userspace?
2710 *
2711 * Return: 0 on success, errno otherwise
2712 */
2713static int tipc_accept(struct socket *sock, struct socket *new_sock, int flags,
2714 bool kern)
2715{
2716 struct sock *new_sk, *sk = sock->sk;
2717 struct tipc_sock *new_tsock;
2718 struct msghdr m = {NULL,};
2719 struct tipc_msg *msg;
2720 struct sk_buff *buf;
2721 long timeo;
2722 int res;
2723
2724 lock_sock(sk);
2725
2726 if (sk->sk_state != TIPC_LISTEN) {
2727 res = -EINVAL;
2728 goto exit;
2729 }
2730 timeo = sock_rcvtimeo(sk, flags & O_NONBLOCK);
2731 res = tipc_wait_for_accept(sock, timeo);
2732 if (res)
2733 goto exit;
2734
2735 buf = skb_peek(&sk->sk_receive_queue);
2736
2737 res = tipc_sk_create(sock_net(sock->sk), new_sock, 0, kern);
2738 if (res)
2739 goto exit;
2740 security_sk_clone(sock->sk, new_sock->sk);
2741
2742 new_sk = new_sock->sk;
2743 new_tsock = tipc_sk(new_sk);
2744 msg = buf_msg(buf);
2745
2746 /* we lock on new_sk; but lockdep sees the lock on sk */
2747 lock_sock_nested(new_sk, SINGLE_DEPTH_NESTING);
2748
2749 /*
2750 * Reject any stray messages received by new socket
2751 * before the socket lock was taken (very, very unlikely)
2752 */
2753 tsk_rej_rx_queue(new_sk, TIPC_ERR_NO_PORT);
2754
2755 /* Connect new socket to it's peer */
2756 tipc_sk_finish_conn(new_tsock, msg_origport(msg), msg_orignode(msg));
2757
2758 tsk_set_importance(new_sk, msg_importance(msg));
2759 if (msg_named(msg)) {
2760 new_tsock->conn_addrtype = TIPC_SERVICE_ADDR;
2761 msg_set_nametype(&new_tsock->phdr, msg_nametype(msg));
2762 msg_set_nameinst(&new_tsock->phdr, msg_nameinst(msg));
2763 }
2764
2765 /*
2766 * Respond to 'SYN-' by discarding it & returning 'ACK'.
2767 * Respond to 'SYN+' by queuing it on new socket & returning 'ACK'.
2768 */
2769 if (!msg_data_sz(msg)) {
2770 tsk_advance_rx_queue(sk);
2771 } else {
2772 __skb_dequeue(&sk->sk_receive_queue);
2773 __skb_queue_head(&new_sk->sk_receive_queue, buf);
2774 skb_set_owner_r(buf, new_sk);
2775 }
2776 __tipc_sendstream(new_sock, &m, 0);
2777 release_sock(new_sk);
2778exit:
2779 release_sock(sk);
2780 return res;
2781}
2782
2783/**
2784 * tipc_shutdown - shutdown socket connection
2785 * @sock: socket structure
2786 * @how: direction to close (must be SHUT_RDWR)
2787 *
2788 * Terminates connection (if necessary), then purges socket's receive queue.
2789 *
2790 * Return: 0 on success, errno otherwise
2791 */
2792static int tipc_shutdown(struct socket *sock, int how)
2793{
2794 struct sock *sk = sock->sk;
2795 int res;
2796
2797 if (how != SHUT_RDWR)
2798 return -EINVAL;
2799
2800 lock_sock(sk);
2801
2802 trace_tipc_sk_shutdown(sk, NULL, TIPC_DUMP_ALL, " ");
2803 __tipc_shutdown(sock, TIPC_CONN_SHUTDOWN);
2804 sk->sk_shutdown = SHUTDOWN_MASK;
2805
2806 if (sk->sk_state == TIPC_DISCONNECTING) {
2807 /* Discard any unreceived messages */
2808 __skb_queue_purge(&sk->sk_receive_queue);
2809
2810 res = 0;
2811 } else {
2812 res = -ENOTCONN;
2813 }
2814 /* Wake up anyone sleeping in poll. */
2815 sk->sk_state_change(sk);
2816
2817 release_sock(sk);
2818 return res;
2819}
2820
2821static void tipc_sk_check_probing_state(struct sock *sk,
2822 struct sk_buff_head *list)
2823{
2824 struct tipc_sock *tsk = tipc_sk(sk);
2825 u32 pnode = tsk_peer_node(tsk);
2826 u32 pport = tsk_peer_port(tsk);
2827 u32 self = tsk_own_node(tsk);
2828 u32 oport = tsk->portid;
2829 struct sk_buff *skb;
2830
2831 if (tsk->probe_unacked) {
2832 tipc_set_sk_state(sk, TIPC_DISCONNECTING);
2833 sk->sk_err = ECONNABORTED;
2834 tipc_node_remove_conn(sock_net(sk), pnode, pport);
2835 sk->sk_state_change(sk);
2836 return;
2837 }
2838 /* Prepare new probe */
2839 skb = tipc_msg_create(CONN_MANAGER, CONN_PROBE, INT_H_SIZE, 0,
2840 pnode, self, pport, oport, TIPC_OK);
2841 if (skb)
2842 __skb_queue_tail(list, skb);
2843 tsk->probe_unacked = true;
2844 sk_reset_timer(sk, &sk->sk_timer, jiffies + CONN_PROBING_INTV);
2845}
2846
2847static void tipc_sk_retry_connect(struct sock *sk, struct sk_buff_head *list)
2848{
2849 struct tipc_sock *tsk = tipc_sk(sk);
2850
2851 /* Try again later if dest link is congested */
2852 if (tsk->cong_link_cnt) {
2853 sk_reset_timer(sk, &sk->sk_timer, msecs_to_jiffies(100));
2854 return;
2855 }
2856 /* Prepare SYN for retransmit */
2857 tipc_msg_skb_clone(&sk->sk_write_queue, list);
2858}
2859
2860static void tipc_sk_timeout(struct timer_list *t)
2861{
2862 struct sock *sk = from_timer(sk, t, sk_timer);
2863 struct tipc_sock *tsk = tipc_sk(sk);
2864 u32 pnode = tsk_peer_node(tsk);
2865 struct sk_buff_head list;
2866 int rc = 0;
2867
2868 __skb_queue_head_init(&list);
2869 bh_lock_sock(sk);
2870
2871 /* Try again later if socket is busy */
2872 if (sock_owned_by_user(sk)) {
2873 sk_reset_timer(sk, &sk->sk_timer, jiffies + HZ / 20);
2874 bh_unlock_sock(sk);
2875 sock_put(sk);
2876 return;
2877 }
2878
2879 if (sk->sk_state == TIPC_ESTABLISHED)
2880 tipc_sk_check_probing_state(sk, &list);
2881 else if (sk->sk_state == TIPC_CONNECTING)
2882 tipc_sk_retry_connect(sk, &list);
2883
2884 bh_unlock_sock(sk);
2885
2886 if (!skb_queue_empty(&list))
2887 rc = tipc_node_xmit(sock_net(sk), &list, pnode, tsk->portid);
2888
2889 /* SYN messages may cause link congestion */
2890 if (rc == -ELINKCONG) {
2891 tipc_dest_push(&tsk->cong_links, pnode, 0);
2892 tsk->cong_link_cnt = 1;
2893 }
2894 sock_put(sk);
2895}
2896
2897static int tipc_sk_publish(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2898{
2899 struct sock *sk = &tsk->sk;
2900 struct net *net = sock_net(sk);
2901 struct tipc_socket_addr skaddr;
2902 struct publication *p;
2903 u32 key;
2904
2905 if (tipc_sk_connected(sk))
2906 return -EINVAL;
2907 key = tsk->portid + tsk->pub_count + 1;
2908 if (key == tsk->portid)
2909 return -EADDRINUSE;
2910 skaddr.ref = tsk->portid;
2911 skaddr.node = tipc_own_addr(net);
2912 p = tipc_nametbl_publish(net, ua, &skaddr, key);
2913 if (unlikely(!p))
2914 return -EINVAL;
2915
2916 list_add(&p->binding_sock, &tsk->publications);
2917 tsk->pub_count++;
2918 tsk->published = true;
2919 return 0;
2920}
2921
2922static int tipc_sk_withdraw(struct tipc_sock *tsk, struct tipc_uaddr *ua)
2923{
2924 struct net *net = sock_net(&tsk->sk);
2925 struct publication *safe, *p;
2926 struct tipc_uaddr _ua;
2927 int rc = -EINVAL;
2928
2929 list_for_each_entry_safe(p, safe, &tsk->publications, binding_sock) {
2930 if (!ua) {
2931 tipc_uaddr(&_ua, TIPC_SERVICE_RANGE, p->scope,
2932 p->sr.type, p->sr.lower, p->sr.upper);
2933 tipc_nametbl_withdraw(net, &_ua, &p->sk, p->key);
2934 continue;
2935 }
2936 /* Unbind specific publication */
2937 if (p->scope != ua->scope)
2938 continue;
2939 if (p->sr.type != ua->sr.type)
2940 continue;
2941 if (p->sr.lower != ua->sr.lower)
2942 continue;
2943 if (p->sr.upper != ua->sr.upper)
2944 break;
2945 tipc_nametbl_withdraw(net, ua, &p->sk, p->key);
2946 rc = 0;
2947 break;
2948 }
2949 if (list_empty(&tsk->publications)) {
2950 tsk->published = 0;
2951 rc = 0;
2952 }
2953 return rc;
2954}
2955
2956/* tipc_sk_reinit: set non-zero address in all existing sockets
2957 * when we go from standalone to network mode.
2958 */
2959void tipc_sk_reinit(struct net *net)
2960{
2961 struct tipc_net *tn = net_generic(net, tipc_net_id);
2962 struct rhashtable_iter iter;
2963 struct tipc_sock *tsk;
2964 struct tipc_msg *msg;
2965
2966 rhashtable_walk_enter(&tn->sk_rht, &iter);
2967
2968 do {
2969 rhashtable_walk_start(&iter);
2970
2971 while ((tsk = rhashtable_walk_next(&iter)) && !IS_ERR(tsk)) {
2972 sock_hold(&tsk->sk);
2973 rhashtable_walk_stop(&iter);
2974 lock_sock(&tsk->sk);
2975 msg = &tsk->phdr;
2976 msg_set_prevnode(msg, tipc_own_addr(net));
2977 msg_set_orignode(msg, tipc_own_addr(net));
2978 release_sock(&tsk->sk);
2979 rhashtable_walk_start(&iter);
2980 sock_put(&tsk->sk);
2981 }
2982
2983 rhashtable_walk_stop(&iter);
2984 } while (tsk == ERR_PTR(-EAGAIN));
2985
2986 rhashtable_walk_exit(&iter);
2987}
2988
2989static struct tipc_sock *tipc_sk_lookup(struct net *net, u32 portid)
2990{
2991 struct tipc_net *tn = net_generic(net, tipc_net_id);
2992 struct tipc_sock *tsk;
2993
2994 rcu_read_lock();
2995 tsk = rhashtable_lookup(&tn->sk_rht, &portid, tsk_rht_params);
2996 if (tsk)
2997 sock_hold(&tsk->sk);
2998 rcu_read_unlock();
2999
3000 return tsk;
3001}
3002
3003static int tipc_sk_insert(struct tipc_sock *tsk)
3004{
3005 struct sock *sk = &tsk->sk;
3006 struct net *net = sock_net(sk);
3007 struct tipc_net *tn = net_generic(net, tipc_net_id);
3008 u32 remaining = (TIPC_MAX_PORT - TIPC_MIN_PORT) + 1;
3009 u32 portid = prandom_u32() % remaining + TIPC_MIN_PORT;
3010
3011 while (remaining--) {
3012 portid++;
3013 if ((portid < TIPC_MIN_PORT) || (portid > TIPC_MAX_PORT))
3014 portid = TIPC_MIN_PORT;
3015 tsk->portid = portid;
3016 sock_hold(&tsk->sk);
3017 if (!rhashtable_lookup_insert_fast(&tn->sk_rht, &tsk->node,
3018 tsk_rht_params))
3019 return 0;
3020 sock_put(&tsk->sk);
3021 }
3022
3023 return -1;
3024}
3025
3026static void tipc_sk_remove(struct tipc_sock *tsk)
3027{
3028 struct sock *sk = &tsk->sk;
3029 struct tipc_net *tn = net_generic(sock_net(sk), tipc_net_id);
3030
3031 if (!rhashtable_remove_fast(&tn->sk_rht, &tsk->node, tsk_rht_params)) {
3032 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
3033 __sock_put(sk);
3034 }
3035}
3036
3037static const struct rhashtable_params tsk_rht_params = {
3038 .nelem_hint = 192,
3039 .head_offset = offsetof(struct tipc_sock, node),
3040 .key_offset = offsetof(struct tipc_sock, portid),
3041 .key_len = sizeof(u32), /* portid */
3042 .max_size = 1048576,
3043 .min_size = 256,
3044 .automatic_shrinking = true,
3045};
3046
3047int tipc_sk_rht_init(struct net *net)
3048{
3049 struct tipc_net *tn = net_generic(net, tipc_net_id);
3050
3051 return rhashtable_init(&tn->sk_rht, &tsk_rht_params);
3052}
3053
3054void tipc_sk_rht_destroy(struct net *net)
3055{
3056 struct tipc_net *tn = net_generic(net, tipc_net_id);
3057
3058 /* Wait for socket readers to complete */
3059 synchronize_net();
3060
3061 rhashtable_destroy(&tn->sk_rht);
3062}
3063
3064static int tipc_sk_join(struct tipc_sock *tsk, struct tipc_group_req *mreq)
3065{
3066 struct net *net = sock_net(&tsk->sk);
3067 struct tipc_group *grp = tsk->group;
3068 struct tipc_msg *hdr = &tsk->phdr;
3069 struct tipc_uaddr ua;
3070 int rc;
3071
3072 if (mreq->type < TIPC_RESERVED_TYPES)
3073 return -EACCES;
3074 if (mreq->scope > TIPC_NODE_SCOPE)
3075 return -EINVAL;
3076 if (mreq->scope != TIPC_NODE_SCOPE)
3077 mreq->scope = TIPC_CLUSTER_SCOPE;
3078 if (grp)
3079 return -EACCES;
3080 grp = tipc_group_create(net, tsk->portid, mreq, &tsk->group_is_open);
3081 if (!grp)
3082 return -ENOMEM;
3083 tsk->group = grp;
3084 msg_set_lookup_scope(hdr, mreq->scope);
3085 msg_set_nametype(hdr, mreq->type);
3086 msg_set_dest_droppable(hdr, true);
3087 tipc_uaddr(&ua, TIPC_SERVICE_RANGE, mreq->scope,
3088 mreq->type, mreq->instance, mreq->instance);
3089 tipc_nametbl_build_group(net, grp, &ua);
3090 rc = tipc_sk_publish(tsk, &ua);
3091 if (rc) {
3092 tipc_group_delete(net, grp);
3093 tsk->group = NULL;
3094 return rc;
3095 }
3096 /* Eliminate any risk that a broadcast overtakes sent JOINs */
3097 tsk->mc_method.rcast = true;
3098 tsk->mc_method.mandatory = true;
3099 tipc_group_join(net, grp, &tsk->sk.sk_rcvbuf);
3100 return rc;
3101}
3102
3103static int tipc_sk_leave(struct tipc_sock *tsk)
3104{
3105 struct net *net = sock_net(&tsk->sk);
3106 struct tipc_group *grp = tsk->group;
3107 struct tipc_uaddr ua;
3108 int scope;
3109
3110 if (!grp)
3111 return -EINVAL;
3112 ua.addrtype = TIPC_SERVICE_RANGE;
3113 tipc_group_self(grp, &ua.sr, &scope);
3114 ua.scope = scope;
3115 tipc_group_delete(net, grp);
3116 tsk->group = NULL;
3117 tipc_sk_withdraw(tsk, &ua);
3118 return 0;
3119}
3120
3121/**
3122 * tipc_setsockopt - set socket option
3123 * @sock: socket structure
3124 * @lvl: option level
3125 * @opt: option identifier
3126 * @ov: pointer to new option value
3127 * @ol: length of option value
3128 *
3129 * For stream sockets only, accepts and ignores all IPPROTO_TCP options
3130 * (to ease compatibility).
3131 *
3132 * Return: 0 on success, errno otherwise
3133 */
3134static int tipc_setsockopt(struct socket *sock, int lvl, int opt,
3135 sockptr_t ov, unsigned int ol)
3136{
3137 struct sock *sk = sock->sk;
3138 struct tipc_sock *tsk = tipc_sk(sk);
3139 struct tipc_group_req mreq;
3140 u32 value = 0;
3141 int res = 0;
3142
3143 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3144 return 0;
3145 if (lvl != SOL_TIPC)
3146 return -ENOPROTOOPT;
3147
3148 switch (opt) {
3149 case TIPC_IMPORTANCE:
3150 case TIPC_SRC_DROPPABLE:
3151 case TIPC_DEST_DROPPABLE:
3152 case TIPC_CONN_TIMEOUT:
3153 case TIPC_NODELAY:
3154 if (ol < sizeof(value))
3155 return -EINVAL;
3156 if (copy_from_sockptr(&value, ov, sizeof(u32)))
3157 return -EFAULT;
3158 break;
3159 case TIPC_GROUP_JOIN:
3160 if (ol < sizeof(mreq))
3161 return -EINVAL;
3162 if (copy_from_sockptr(&mreq, ov, sizeof(mreq)))
3163 return -EFAULT;
3164 break;
3165 default:
3166 if (!sockptr_is_null(ov) || ol)
3167 return -EINVAL;
3168 }
3169
3170 lock_sock(sk);
3171
3172 switch (opt) {
3173 case TIPC_IMPORTANCE:
3174 res = tsk_set_importance(sk, value);
3175 break;
3176 case TIPC_SRC_DROPPABLE:
3177 if (sock->type != SOCK_STREAM)
3178 tsk_set_unreliable(tsk, value);
3179 else
3180 res = -ENOPROTOOPT;
3181 break;
3182 case TIPC_DEST_DROPPABLE:
3183 tsk_set_unreturnable(tsk, value);
3184 break;
3185 case TIPC_CONN_TIMEOUT:
3186 tipc_sk(sk)->conn_timeout = value;
3187 break;
3188 case TIPC_MCAST_BROADCAST:
3189 tsk->mc_method.rcast = false;
3190 tsk->mc_method.mandatory = true;
3191 break;
3192 case TIPC_MCAST_REPLICAST:
3193 tsk->mc_method.rcast = true;
3194 tsk->mc_method.mandatory = true;
3195 break;
3196 case TIPC_GROUP_JOIN:
3197 res = tipc_sk_join(tsk, &mreq);
3198 break;
3199 case TIPC_GROUP_LEAVE:
3200 res = tipc_sk_leave(tsk);
3201 break;
3202 case TIPC_NODELAY:
3203 tsk->nodelay = !!value;
3204 tsk_set_nagle(tsk);
3205 break;
3206 default:
3207 res = -EINVAL;
3208 }
3209
3210 release_sock(sk);
3211
3212 return res;
3213}
3214
3215/**
3216 * tipc_getsockopt - get socket option
3217 * @sock: socket structure
3218 * @lvl: option level
3219 * @opt: option identifier
3220 * @ov: receptacle for option value
3221 * @ol: receptacle for length of option value
3222 *
3223 * For stream sockets only, returns 0 length result for all IPPROTO_TCP options
3224 * (to ease compatibility).
3225 *
3226 * Return: 0 on success, errno otherwise
3227 */
3228static int tipc_getsockopt(struct socket *sock, int lvl, int opt,
3229 char __user *ov, int __user *ol)
3230{
3231 struct sock *sk = sock->sk;
3232 struct tipc_sock *tsk = tipc_sk(sk);
3233 struct tipc_service_range seq;
3234 int len, scope;
3235 u32 value;
3236 int res;
3237
3238 if ((lvl == IPPROTO_TCP) && (sock->type == SOCK_STREAM))
3239 return put_user(0, ol);
3240 if (lvl != SOL_TIPC)
3241 return -ENOPROTOOPT;
3242 res = get_user(len, ol);
3243 if (res)
3244 return res;
3245
3246 lock_sock(sk);
3247
3248 switch (opt) {
3249 case TIPC_IMPORTANCE:
3250 value = tsk_importance(tsk);
3251 break;
3252 case TIPC_SRC_DROPPABLE:
3253 value = tsk_unreliable(tsk);
3254 break;
3255 case TIPC_DEST_DROPPABLE:
3256 value = tsk_unreturnable(tsk);
3257 break;
3258 case TIPC_CONN_TIMEOUT:
3259 value = tsk->conn_timeout;
3260 /* no need to set "res", since already 0 at this point */
3261 break;
3262 case TIPC_NODE_RECVQ_DEPTH:
3263 value = 0; /* was tipc_queue_size, now obsolete */
3264 break;
3265 case TIPC_SOCK_RECVQ_DEPTH:
3266 value = skb_queue_len(&sk->sk_receive_queue);
3267 break;
3268 case TIPC_SOCK_RECVQ_USED:
3269 value = sk_rmem_alloc_get(sk);
3270 break;
3271 case TIPC_GROUP_JOIN:
3272 seq.type = 0;
3273 if (tsk->group)
3274 tipc_group_self(tsk->group, &seq, &scope);
3275 value = seq.type;
3276 break;
3277 default:
3278 res = -EINVAL;
3279 }
3280
3281 release_sock(sk);
3282
3283 if (res)
3284 return res; /* "get" failed */
3285
3286 if (len < sizeof(value))
3287 return -EINVAL;
3288
3289 if (copy_to_user(ov, &value, sizeof(value)))
3290 return -EFAULT;
3291
3292 return put_user(sizeof(value), ol);
3293}
3294
3295static int tipc_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
3296{
3297 struct net *net = sock_net(sock->sk);
3298 struct tipc_sioc_nodeid_req nr = {0};
3299 struct tipc_sioc_ln_req lnr;
3300 void __user *argp = (void __user *)arg;
3301
3302 switch (cmd) {
3303 case SIOCGETLINKNAME:
3304 if (copy_from_user(&lnr, argp, sizeof(lnr)))
3305 return -EFAULT;
3306 if (!tipc_node_get_linkname(net,
3307 lnr.bearer_id & 0xffff, lnr.peer,
3308 lnr.linkname, TIPC_MAX_LINK_NAME)) {
3309 if (copy_to_user(argp, &lnr, sizeof(lnr)))
3310 return -EFAULT;
3311 return 0;
3312 }
3313 return -EADDRNOTAVAIL;
3314 case SIOCGETNODEID:
3315 if (copy_from_user(&nr, argp, sizeof(nr)))
3316 return -EFAULT;
3317 if (!tipc_node_get_id(net, nr.peer, nr.node_id))
3318 return -EADDRNOTAVAIL;
3319 if (copy_to_user(argp, &nr, sizeof(nr)))
3320 return -EFAULT;
3321 return 0;
3322 default:
3323 return -ENOIOCTLCMD;
3324 }
3325}
3326
3327static int tipc_socketpair(struct socket *sock1, struct socket *sock2)
3328{
3329 struct tipc_sock *tsk2 = tipc_sk(sock2->sk);
3330 struct tipc_sock *tsk1 = tipc_sk(sock1->sk);
3331 u32 onode = tipc_own_addr(sock_net(sock1->sk));
3332
3333 tsk1->peer.family = AF_TIPC;
3334 tsk1->peer.addrtype = TIPC_SOCKET_ADDR;
3335 tsk1->peer.scope = TIPC_NODE_SCOPE;
3336 tsk1->peer.addr.id.ref = tsk2->portid;
3337 tsk1->peer.addr.id.node = onode;
3338 tsk2->peer.family = AF_TIPC;
3339 tsk2->peer.addrtype = TIPC_SOCKET_ADDR;
3340 tsk2->peer.scope = TIPC_NODE_SCOPE;
3341 tsk2->peer.addr.id.ref = tsk1->portid;
3342 tsk2->peer.addr.id.node = onode;
3343
3344 tipc_sk_finish_conn(tsk1, tsk2->portid, onode);
3345 tipc_sk_finish_conn(tsk2, tsk1->portid, onode);
3346 return 0;
3347}
3348
3349/* Protocol switches for the various types of TIPC sockets */
3350
3351static const struct proto_ops msg_ops = {
3352 .owner = THIS_MODULE,
3353 .family = AF_TIPC,
3354 .release = tipc_release,
3355 .bind = tipc_bind,
3356 .connect = tipc_connect,
3357 .socketpair = tipc_socketpair,
3358 .accept = sock_no_accept,
3359 .getname = tipc_getname,
3360 .poll = tipc_poll,
3361 .ioctl = tipc_ioctl,
3362 .listen = sock_no_listen,
3363 .shutdown = tipc_shutdown,
3364 .setsockopt = tipc_setsockopt,
3365 .getsockopt = tipc_getsockopt,
3366 .sendmsg = tipc_sendmsg,
3367 .recvmsg = tipc_recvmsg,
3368 .mmap = sock_no_mmap,
3369 .sendpage = sock_no_sendpage
3370};
3371
3372static const struct proto_ops packet_ops = {
3373 .owner = THIS_MODULE,
3374 .family = AF_TIPC,
3375 .release = tipc_release,
3376 .bind = tipc_bind,
3377 .connect = tipc_connect,
3378 .socketpair = tipc_socketpair,
3379 .accept = tipc_accept,
3380 .getname = tipc_getname,
3381 .poll = tipc_poll,
3382 .ioctl = tipc_ioctl,
3383 .listen = tipc_listen,
3384 .shutdown = tipc_shutdown,
3385 .setsockopt = tipc_setsockopt,
3386 .getsockopt = tipc_getsockopt,
3387 .sendmsg = tipc_send_packet,
3388 .recvmsg = tipc_recvmsg,
3389 .mmap = sock_no_mmap,
3390 .sendpage = sock_no_sendpage
3391};
3392
3393static const struct proto_ops stream_ops = {
3394 .owner = THIS_MODULE,
3395 .family = AF_TIPC,
3396 .release = tipc_release,
3397 .bind = tipc_bind,
3398 .connect = tipc_connect,
3399 .socketpair = tipc_socketpair,
3400 .accept = tipc_accept,
3401 .getname = tipc_getname,
3402 .poll = tipc_poll,
3403 .ioctl = tipc_ioctl,
3404 .listen = tipc_listen,
3405 .shutdown = tipc_shutdown,
3406 .setsockopt = tipc_setsockopt,
3407 .getsockopt = tipc_getsockopt,
3408 .sendmsg = tipc_sendstream,
3409 .recvmsg = tipc_recvstream,
3410 .mmap = sock_no_mmap,
3411 .sendpage = sock_no_sendpage
3412};
3413
3414static const struct net_proto_family tipc_family_ops = {
3415 .owner = THIS_MODULE,
3416 .family = AF_TIPC,
3417 .create = tipc_sk_create
3418};
3419
3420static struct proto tipc_proto = {
3421 .name = "TIPC",
3422 .owner = THIS_MODULE,
3423 .obj_size = sizeof(struct tipc_sock),
3424 .sysctl_rmem = sysctl_tipc_rmem
3425};
3426
3427/**
3428 * tipc_socket_init - initialize TIPC socket interface
3429 *
3430 * Return: 0 on success, errno otherwise
3431 */
3432int tipc_socket_init(void)
3433{
3434 int res;
3435
3436 res = proto_register(&tipc_proto, 1);
3437 if (res) {
3438 pr_err("Failed to register TIPC protocol type\n");
3439 goto out;
3440 }
3441
3442 res = sock_register(&tipc_family_ops);
3443 if (res) {
3444 pr_err("Failed to register TIPC socket type\n");
3445 proto_unregister(&tipc_proto);
3446 goto out;
3447 }
3448 out:
3449 return res;
3450}
3451
3452/**
3453 * tipc_socket_stop - stop TIPC socket interface
3454 */
3455void tipc_socket_stop(void)
3456{
3457 sock_unregister(tipc_family_ops.family);
3458 proto_unregister(&tipc_proto);
3459}
3460
3461/* Caller should hold socket lock for the passed tipc socket. */
3462static int __tipc_nl_add_sk_con(struct sk_buff *skb, struct tipc_sock *tsk)
3463{
3464 u32 peer_node, peer_port;
3465 u32 conn_type, conn_instance;
3466 struct nlattr *nest;
3467
3468 peer_node = tsk_peer_node(tsk);
3469 peer_port = tsk_peer_port(tsk);
3470 conn_type = msg_nametype(&tsk->phdr);
3471 conn_instance = msg_nameinst(&tsk->phdr);
3472 nest = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_CON);
3473 if (!nest)
3474 return -EMSGSIZE;
3475
3476 if (nla_put_u32(skb, TIPC_NLA_CON_NODE, peer_node))
3477 goto msg_full;
3478 if (nla_put_u32(skb, TIPC_NLA_CON_SOCK, peer_port))
3479 goto msg_full;
3480
3481 if (tsk->conn_addrtype != 0) {
3482 if (nla_put_flag(skb, TIPC_NLA_CON_FLAG))
3483 goto msg_full;
3484 if (nla_put_u32(skb, TIPC_NLA_CON_TYPE, conn_type))
3485 goto msg_full;
3486 if (nla_put_u32(skb, TIPC_NLA_CON_INST, conn_instance))
3487 goto msg_full;
3488 }
3489 nla_nest_end(skb, nest);
3490
3491 return 0;
3492
3493msg_full:
3494 nla_nest_cancel(skb, nest);
3495
3496 return -EMSGSIZE;
3497}
3498
3499static int __tipc_nl_add_sk_info(struct sk_buff *skb, struct tipc_sock
3500 *tsk)
3501{
3502 struct net *net = sock_net(skb->sk);
3503 struct sock *sk = &tsk->sk;
3504
3505 if (nla_put_u32(skb, TIPC_NLA_SOCK_REF, tsk->portid) ||
3506 nla_put_u32(skb, TIPC_NLA_SOCK_ADDR, tipc_own_addr(net)))
3507 return -EMSGSIZE;
3508
3509 if (tipc_sk_connected(sk)) {
3510 if (__tipc_nl_add_sk_con(skb, tsk))
3511 return -EMSGSIZE;
3512 } else if (!list_empty(&tsk->publications)) {
3513 if (nla_put_flag(skb, TIPC_NLA_SOCK_HAS_PUBL))
3514 return -EMSGSIZE;
3515 }
3516 return 0;
3517}
3518
3519/* Caller should hold socket lock for the passed tipc socket. */
3520static int __tipc_nl_add_sk(struct sk_buff *skb, struct netlink_callback *cb,
3521 struct tipc_sock *tsk)
3522{
3523 struct nlattr *attrs;
3524 void *hdr;
3525
3526 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3527 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_SOCK_GET);
3528 if (!hdr)
3529 goto msg_cancel;
3530
3531 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3532 if (!attrs)
3533 goto genlmsg_cancel;
3534
3535 if (__tipc_nl_add_sk_info(skb, tsk))
3536 goto attr_msg_cancel;
3537
3538 nla_nest_end(skb, attrs);
3539 genlmsg_end(skb, hdr);
3540
3541 return 0;
3542
3543attr_msg_cancel:
3544 nla_nest_cancel(skb, attrs);
3545genlmsg_cancel:
3546 genlmsg_cancel(skb, hdr);
3547msg_cancel:
3548 return -EMSGSIZE;
3549}
3550
3551int tipc_nl_sk_walk(struct sk_buff *skb, struct netlink_callback *cb,
3552 int (*skb_handler)(struct sk_buff *skb,
3553 struct netlink_callback *cb,
3554 struct tipc_sock *tsk))
3555{
3556 struct rhashtable_iter *iter = (void *)cb->args[4];
3557 struct tipc_sock *tsk;
3558 int err;
3559
3560 rhashtable_walk_start(iter);
3561 while ((tsk = rhashtable_walk_next(iter)) != NULL) {
3562 if (IS_ERR(tsk)) {
3563 err = PTR_ERR(tsk);
3564 if (err == -EAGAIN) {
3565 err = 0;
3566 continue;
3567 }
3568 break;
3569 }
3570
3571 sock_hold(&tsk->sk);
3572 rhashtable_walk_stop(iter);
3573 lock_sock(&tsk->sk);
3574 err = skb_handler(skb, cb, tsk);
3575 if (err) {
3576 release_sock(&tsk->sk);
3577 sock_put(&tsk->sk);
3578 goto out;
3579 }
3580 release_sock(&tsk->sk);
3581 rhashtable_walk_start(iter);
3582 sock_put(&tsk->sk);
3583 }
3584 rhashtable_walk_stop(iter);
3585out:
3586 return skb->len;
3587}
3588EXPORT_SYMBOL(tipc_nl_sk_walk);
3589
3590int tipc_dump_start(struct netlink_callback *cb)
3591{
3592 return __tipc_dump_start(cb, sock_net(cb->skb->sk));
3593}
3594EXPORT_SYMBOL(tipc_dump_start);
3595
3596int __tipc_dump_start(struct netlink_callback *cb, struct net *net)
3597{
3598 /* tipc_nl_name_table_dump() uses cb->args[0...3]. */
3599 struct rhashtable_iter *iter = (void *)cb->args[4];
3600 struct tipc_net *tn = tipc_net(net);
3601
3602 if (!iter) {
3603 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
3604 if (!iter)
3605 return -ENOMEM;
3606
3607 cb->args[4] = (long)iter;
3608 }
3609
3610 rhashtable_walk_enter(&tn->sk_rht, iter);
3611 return 0;
3612}
3613
3614int tipc_dump_done(struct netlink_callback *cb)
3615{
3616 struct rhashtable_iter *hti = (void *)cb->args[4];
3617
3618 rhashtable_walk_exit(hti);
3619 kfree(hti);
3620 return 0;
3621}
3622EXPORT_SYMBOL(tipc_dump_done);
3623
3624int tipc_sk_fill_sock_diag(struct sk_buff *skb, struct netlink_callback *cb,
3625 struct tipc_sock *tsk, u32 sk_filter_state,
3626 u64 (*tipc_diag_gen_cookie)(struct sock *sk))
3627{
3628 struct sock *sk = &tsk->sk;
3629 struct nlattr *attrs;
3630 struct nlattr *stat;
3631
3632 /*filter response w.r.t sk_state*/
3633 if (!(sk_filter_state & (1 << sk->sk_state)))
3634 return 0;
3635
3636 attrs = nla_nest_start_noflag(skb, TIPC_NLA_SOCK);
3637 if (!attrs)
3638 goto msg_cancel;
3639
3640 if (__tipc_nl_add_sk_info(skb, tsk))
3641 goto attr_msg_cancel;
3642
3643 if (nla_put_u32(skb, TIPC_NLA_SOCK_TYPE, (u32)sk->sk_type) ||
3644 nla_put_u32(skb, TIPC_NLA_SOCK_TIPC_STATE, (u32)sk->sk_state) ||
3645 nla_put_u32(skb, TIPC_NLA_SOCK_INO, sock_i_ino(sk)) ||
3646 nla_put_u32(skb, TIPC_NLA_SOCK_UID,
3647 from_kuid_munged(sk_user_ns(NETLINK_CB(cb->skb).sk),
3648 sock_i_uid(sk))) ||
3649 nla_put_u64_64bit(skb, TIPC_NLA_SOCK_COOKIE,
3650 tipc_diag_gen_cookie(sk),
3651 TIPC_NLA_SOCK_PAD))
3652 goto attr_msg_cancel;
3653
3654 stat = nla_nest_start_noflag(skb, TIPC_NLA_SOCK_STAT);
3655 if (!stat)
3656 goto attr_msg_cancel;
3657
3658 if (nla_put_u32(skb, TIPC_NLA_SOCK_STAT_RCVQ,
3659 skb_queue_len(&sk->sk_receive_queue)) ||
3660 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_SENDQ,
3661 skb_queue_len(&sk->sk_write_queue)) ||
3662 nla_put_u32(skb, TIPC_NLA_SOCK_STAT_DROP,
3663 atomic_read(&sk->sk_drops)))
3664 goto stat_msg_cancel;
3665
3666 if (tsk->cong_link_cnt &&
3667 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_LINK_CONG))
3668 goto stat_msg_cancel;
3669
3670 if (tsk_conn_cong(tsk) &&
3671 nla_put_flag(skb, TIPC_NLA_SOCK_STAT_CONN_CONG))
3672 goto stat_msg_cancel;
3673
3674 nla_nest_end(skb, stat);
3675
3676 if (tsk->group)
3677 if (tipc_group_fill_sock_diag(tsk->group, skb))
3678 goto stat_msg_cancel;
3679
3680 nla_nest_end(skb, attrs);
3681
3682 return 0;
3683
3684stat_msg_cancel:
3685 nla_nest_cancel(skb, stat);
3686attr_msg_cancel:
3687 nla_nest_cancel(skb, attrs);
3688msg_cancel:
3689 return -EMSGSIZE;
3690}
3691EXPORT_SYMBOL(tipc_sk_fill_sock_diag);
3692
3693int tipc_nl_sk_dump(struct sk_buff *skb, struct netlink_callback *cb)
3694{
3695 return tipc_nl_sk_walk(skb, cb, __tipc_nl_add_sk);
3696}
3697
3698/* Caller should hold socket lock for the passed tipc socket. */
3699static int __tipc_nl_add_sk_publ(struct sk_buff *skb,
3700 struct netlink_callback *cb,
3701 struct publication *publ)
3702{
3703 void *hdr;
3704 struct nlattr *attrs;
3705
3706 hdr = genlmsg_put(skb, NETLINK_CB(cb->skb).portid, cb->nlh->nlmsg_seq,
3707 &tipc_genl_family, NLM_F_MULTI, TIPC_NL_PUBL_GET);
3708 if (!hdr)
3709 goto msg_cancel;
3710
3711 attrs = nla_nest_start_noflag(skb, TIPC_NLA_PUBL);
3712 if (!attrs)
3713 goto genlmsg_cancel;
3714
3715 if (nla_put_u32(skb, TIPC_NLA_PUBL_KEY, publ->key))
3716 goto attr_msg_cancel;
3717 if (nla_put_u32(skb, TIPC_NLA_PUBL_TYPE, publ->sr.type))
3718 goto attr_msg_cancel;
3719 if (nla_put_u32(skb, TIPC_NLA_PUBL_LOWER, publ->sr.lower))
3720 goto attr_msg_cancel;
3721 if (nla_put_u32(skb, TIPC_NLA_PUBL_UPPER, publ->sr.upper))
3722 goto attr_msg_cancel;
3723
3724 nla_nest_end(skb, attrs);
3725 genlmsg_end(skb, hdr);
3726
3727 return 0;
3728
3729attr_msg_cancel:
3730 nla_nest_cancel(skb, attrs);
3731genlmsg_cancel:
3732 genlmsg_cancel(skb, hdr);
3733msg_cancel:
3734 return -EMSGSIZE;
3735}
3736
3737/* Caller should hold socket lock for the passed tipc socket. */
3738static int __tipc_nl_list_sk_publ(struct sk_buff *skb,
3739 struct netlink_callback *cb,
3740 struct tipc_sock *tsk, u32 *last_publ)
3741{
3742 int err;
3743 struct publication *p;
3744
3745 if (*last_publ) {
3746 list_for_each_entry(p, &tsk->publications, binding_sock) {
3747 if (p->key == *last_publ)
3748 break;
3749 }
3750 if (p->key != *last_publ) {
3751 /* We never set seq or call nl_dump_check_consistent()
3752 * this means that setting prev_seq here will cause the
3753 * consistence check to fail in the netlink callback
3754 * handler. Resulting in the last NLMSG_DONE message
3755 * having the NLM_F_DUMP_INTR flag set.
3756 */
3757 cb->prev_seq = 1;
3758 *last_publ = 0;
3759 return -EPIPE;
3760 }
3761 } else {
3762 p = list_first_entry(&tsk->publications, struct publication,
3763 binding_sock);
3764 }
3765
3766 list_for_each_entry_from(p, &tsk->publications, binding_sock) {
3767 err = __tipc_nl_add_sk_publ(skb, cb, p);
3768 if (err) {
3769 *last_publ = p->key;
3770 return err;
3771 }
3772 }
3773 *last_publ = 0;
3774
3775 return 0;
3776}
3777
3778int tipc_nl_publ_dump(struct sk_buff *skb, struct netlink_callback *cb)
3779{
3780 int err;
3781 u32 tsk_portid = cb->args[0];
3782 u32 last_publ = cb->args[1];
3783 u32 done = cb->args[2];
3784 struct net *net = sock_net(skb->sk);
3785 struct tipc_sock *tsk;
3786
3787 if (!tsk_portid) {
3788 struct nlattr **attrs = genl_dumpit_info(cb)->attrs;
3789 struct nlattr *sock[TIPC_NLA_SOCK_MAX + 1];
3790
3791 if (!attrs[TIPC_NLA_SOCK])
3792 return -EINVAL;
3793
3794 err = nla_parse_nested_deprecated(sock, TIPC_NLA_SOCK_MAX,
3795 attrs[TIPC_NLA_SOCK],
3796 tipc_nl_sock_policy, NULL);
3797 if (err)
3798 return err;
3799
3800 if (!sock[TIPC_NLA_SOCK_REF])
3801 return -EINVAL;
3802
3803 tsk_portid = nla_get_u32(sock[TIPC_NLA_SOCK_REF]);
3804 }
3805
3806 if (done)
3807 return 0;
3808
3809 tsk = tipc_sk_lookup(net, tsk_portid);
3810 if (!tsk)
3811 return -EINVAL;
3812
3813 lock_sock(&tsk->sk);
3814 err = __tipc_nl_list_sk_publ(skb, cb, tsk, &last_publ);
3815 if (!err)
3816 done = 1;
3817 release_sock(&tsk->sk);
3818 sock_put(&tsk->sk);
3819
3820 cb->args[0] = tsk_portid;
3821 cb->args[1] = last_publ;
3822 cb->args[2] = done;
3823
3824 return skb->len;
3825}
3826
3827/**
3828 * tipc_sk_filtering - check if a socket should be traced
3829 * @sk: the socket to be examined
3830 *
3831 * @sysctl_tipc_sk_filter is used as the socket tuple for filtering:
3832 * (portid, sock type, name type, name lower, name upper)
3833 *
3834 * Return: true if the socket meets the socket tuple data
3835 * (value 0 = 'any') or when there is no tuple set (all = 0),
3836 * otherwise false
3837 */
3838bool tipc_sk_filtering(struct sock *sk)
3839{
3840 struct tipc_sock *tsk;
3841 struct publication *p;
3842 u32 _port, _sktype, _type, _lower, _upper;
3843 u32 type = 0, lower = 0, upper = 0;
3844
3845 if (!sk)
3846 return true;
3847
3848 tsk = tipc_sk(sk);
3849
3850 _port = sysctl_tipc_sk_filter[0];
3851 _sktype = sysctl_tipc_sk_filter[1];
3852 _type = sysctl_tipc_sk_filter[2];
3853 _lower = sysctl_tipc_sk_filter[3];
3854 _upper = sysctl_tipc_sk_filter[4];
3855
3856 if (!_port && !_sktype && !_type && !_lower && !_upper)
3857 return true;
3858
3859 if (_port)
3860 return (_port == tsk->portid);
3861
3862 if (_sktype && _sktype != sk->sk_type)
3863 return false;
3864
3865 if (tsk->published) {
3866 p = list_first_entry_or_null(&tsk->publications,
3867 struct publication, binding_sock);
3868 if (p) {
3869 type = p->sr.type;
3870 lower = p->sr.lower;
3871 upper = p->sr.upper;
3872 }
3873 }
3874
3875 if (!tipc_sk_type_connectionless(sk)) {
3876 type = msg_nametype(&tsk->phdr);
3877 lower = msg_nameinst(&tsk->phdr);
3878 upper = lower;
3879 }
3880
3881 if ((_type && _type != type) || (_lower && _lower != lower) ||
3882 (_upper && _upper != upper))
3883 return false;
3884
3885 return true;
3886}
3887
3888u32 tipc_sock_get_portid(struct sock *sk)
3889{
3890 return (sk) ? (tipc_sk(sk))->portid : 0;
3891}
3892
3893/**
3894 * tipc_sk_overlimit1 - check if socket rx queue is about to be overloaded,
3895 * both the rcv and backlog queues are considered
3896 * @sk: tipc sk to be checked
3897 * @skb: tipc msg to be checked
3898 *
3899 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3900 */
3901
3902bool tipc_sk_overlimit1(struct sock *sk, struct sk_buff *skb)
3903{
3904 atomic_t *dcnt = &tipc_sk(sk)->dupl_rcvcnt;
3905 unsigned int lim = rcvbuf_limit(sk, skb) + atomic_read(dcnt);
3906 unsigned int qsize = sk->sk_backlog.len + sk_rmem_alloc_get(sk);
3907
3908 return (qsize > lim * 90 / 100);
3909}
3910
3911/**
3912 * tipc_sk_overlimit2 - check if socket rx queue is about to be overloaded,
3913 * only the rcv queue is considered
3914 * @sk: tipc sk to be checked
3915 * @skb: tipc msg to be checked
3916 *
3917 * Return: true if the socket rx queue allocation is > 90%, otherwise false
3918 */
3919
3920bool tipc_sk_overlimit2(struct sock *sk, struct sk_buff *skb)
3921{
3922 unsigned int lim = rcvbuf_limit(sk, skb);
3923 unsigned int qsize = sk_rmem_alloc_get(sk);
3924
3925 return (qsize > lim * 90 / 100);
3926}
3927
3928/**
3929 * tipc_sk_dump - dump TIPC socket
3930 * @sk: tipc sk to be dumped
3931 * @dqueues: bitmask to decide if any socket queue to be dumped?
3932 * - TIPC_DUMP_NONE: don't dump socket queues
3933 * - TIPC_DUMP_SK_SNDQ: dump socket send queue
3934 * - TIPC_DUMP_SK_RCVQ: dump socket rcv queue
3935 * - TIPC_DUMP_SK_BKLGQ: dump socket backlog queue
3936 * - TIPC_DUMP_ALL: dump all the socket queues above
3937 * @buf: returned buffer of dump data in format
3938 */
3939int tipc_sk_dump(struct sock *sk, u16 dqueues, char *buf)
3940{
3941 int i = 0;
3942 size_t sz = (dqueues) ? SK_LMAX : SK_LMIN;
3943 u32 conn_type, conn_instance;
3944 struct tipc_sock *tsk;
3945 struct publication *p;
3946 bool tsk_connected;
3947
3948 if (!sk) {
3949 i += scnprintf(buf, sz, "sk data: (null)\n");
3950 return i;
3951 }
3952
3953 tsk = tipc_sk(sk);
3954 tsk_connected = !tipc_sk_type_connectionless(sk);
3955
3956 i += scnprintf(buf, sz, "sk data: %u", sk->sk_type);
3957 i += scnprintf(buf + i, sz - i, " %d", sk->sk_state);
3958 i += scnprintf(buf + i, sz - i, " %x", tsk_own_node(tsk));
3959 i += scnprintf(buf + i, sz - i, " %u", tsk->portid);
3960 i += scnprintf(buf + i, sz - i, " | %u", tsk_connected);
3961 if (tsk_connected) {
3962 i += scnprintf(buf + i, sz - i, " %x", tsk_peer_node(tsk));
3963 i += scnprintf(buf + i, sz - i, " %u", tsk_peer_port(tsk));
3964 conn_type = msg_nametype(&tsk->phdr);
3965 conn_instance = msg_nameinst(&tsk->phdr);
3966 i += scnprintf(buf + i, sz - i, " %u", conn_type);
3967 i += scnprintf(buf + i, sz - i, " %u", conn_instance);
3968 }
3969 i += scnprintf(buf + i, sz - i, " | %u", tsk->published);
3970 if (tsk->published) {
3971 p = list_first_entry_or_null(&tsk->publications,
3972 struct publication, binding_sock);
3973 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.type : 0);
3974 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.lower : 0);
3975 i += scnprintf(buf + i, sz - i, " %u", (p) ? p->sr.upper : 0);
3976 }
3977 i += scnprintf(buf + i, sz - i, " | %u", tsk->snd_win);
3978 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_win);
3979 i += scnprintf(buf + i, sz - i, " %u", tsk->max_pkt);
3980 i += scnprintf(buf + i, sz - i, " %x", tsk->peer_caps);
3981 i += scnprintf(buf + i, sz - i, " %u", tsk->cong_link_cnt);
3982 i += scnprintf(buf + i, sz - i, " %u", tsk->snt_unacked);
3983 i += scnprintf(buf + i, sz - i, " %u", tsk->rcv_unacked);
3984 i += scnprintf(buf + i, sz - i, " %u", atomic_read(&tsk->dupl_rcvcnt));
3985 i += scnprintf(buf + i, sz - i, " %u", sk->sk_shutdown);
3986 i += scnprintf(buf + i, sz - i, " | %d", sk_wmem_alloc_get(sk));
3987 i += scnprintf(buf + i, sz - i, " %d", sk->sk_sndbuf);
3988 i += scnprintf(buf + i, sz - i, " | %d", sk_rmem_alloc_get(sk));
3989 i += scnprintf(buf + i, sz - i, " %d", sk->sk_rcvbuf);
3990 i += scnprintf(buf + i, sz - i, " | %d\n", READ_ONCE(sk->sk_backlog.len));
3991
3992 if (dqueues & TIPC_DUMP_SK_SNDQ) {
3993 i += scnprintf(buf + i, sz - i, "sk_write_queue: ");
3994 i += tipc_list_dump(&sk->sk_write_queue, false, buf + i);
3995 }
3996
3997 if (dqueues & TIPC_DUMP_SK_RCVQ) {
3998 i += scnprintf(buf + i, sz - i, "sk_receive_queue: ");
3999 i += tipc_list_dump(&sk->sk_receive_queue, false, buf + i);
4000 }
4001
4002 if (dqueues & TIPC_DUMP_SK_BKLGQ) {
4003 i += scnprintf(buf + i, sz - i, "sk_backlog:\n head ");
4004 i += tipc_skb_dump(sk->sk_backlog.head, false, buf + i);
4005 if (sk->sk_backlog.tail != sk->sk_backlog.head) {
4006 i += scnprintf(buf + i, sz - i, " tail ");
4007 i += tipc_skb_dump(sk->sk_backlog.tail, false,
4008 buf + i);
4009 }
4010 }
4011
4012 return i;
4013}